JP2023167643A - Robot operation plan supporting system, robot operation plan supporting method, and computer program - Google Patents

Abstract

To support an operator so as to enable the operator to select an efficient procedure when a robot performs specific work, more easily than a conventional one.SOLUTION: Pieces of procedure data indicating a plurality of procedures when a robot performs specific work are given to a computer. The pieces of procedure data are given by an operator manually operating the robot, for example. The computer generates a Petri net 7 according to α algorithm on the basis of the pieces of procedure data.SELECTED DRAWING: Figure 7

Description

TECHNICAL FIELD The present invention relates to technology for supporting motion planning when having a robot perform a task.

Conventionally, robots such as industrial robots have been used to rapidly produce products such as automobiles, machinery, electrical equipment, electronic equipment, foods, cosmetics, and pharmaceuticals in large quantities and at high speed. Robots have been used to save labor and labor, but a wide variety of products are being developed one after another to meet the needs of consumers, and the cost of introducing specialized robots for each product increases.

Therefore, versatile robots have been introduced to perform tasks tailored to each product. In this case, a method called ``teaching playback'' is widely adopted in which a human (operator) teaches the robot the work procedure and prepares a program by touching the robot directly or moving it with an operating terminal. Teaching playback is sometimes called "online teaching" or "playback-style teaching."

However, when operating a robot, it is necessary to specify a program in advance according to the product, and the robot only operates according to the program. That is, it is not easy for a robot to autonomously select the optimal operation depending on the situation.

Furthermore, the more complicated the work process is, the more difficult it becomes for the operator to teach it. That is, teaching and playback must be performed by a skilled operator.

Therefore, systems for controlling robots using Petri nets have been proposed, as described in Non-Patent Documents 1 and 2. A Petri net is a graphical model of the flow of work performed by a robot. Therefore, by using Petri nets, it is possible to facilitate the control of robots.

Genichi Yasuda, “Control of industrial robot systems based on Petri nets”, 51st Automatic Control Joint Conference pp.974-975, (2008), https://www.jstage.jst.go.jp/article /jacc/51/0/51_0_213/_pdf Yosuke Funami, Toru Kudo, Keiji Watanabe, “Control of robot systems using Petri nets”, 170th Research Meeting of the Tohoku Branch of the Society of Instrument and Control Engineers, pp. 1-8, (1997), https://www.topic .ad.jp/sice/htdocs/papers/170/170-6.pdf

By the way, even one specific task can take many different steps. For example, as a procedure for moving a first workpiece and a second workpiece that are in separate positions to a predetermined position, the first workpiece is moved to a predetermined position, and then the second workpiece is moved to a predetermined position. There are multiple steps, such as moving the workpiece, placing the first workpiece on top of the second workpiece, and then moving both workpieces together to a predetermined position.

In order to have the robot perform the work as efficiently as possible, it is necessary to select a procedure that is as close to the optimum as possible.

In view of these problems, it is an object of the present invention to assist an operator in selecting an efficient procedure for having a robot perform a specific task more easily than in the past.

A robot motion planning support system according to an embodiment of the present invention includes a procedure data acquisition unit that acquires procedure data indicating each of a plurality of procedures for causing a robot to perform a specific task; and process model generation means for generating a process model of the work based on the procedure data.

Preferably, a route that satisfies predetermined requirements from all or part of a plurality of routes that can be taken from an initial node to a target node in the process model is selected as a qualified route suitable for the work. route selection means.

Alternatively, the eligible route selection means randomly selects N routes from among the plurality of routes, and selects the route that requires the least time and cost for the robot among the N routes to be the eligible route. Alternatively, the process model generating means generates a Petri net as the process model using the α algorithm, and the motion plan determining means randomly advances tokens in places having branches in the Petri net. By doing so, multiple routes and timings are generated, and an eligible route is selected from these routes.

According to the present invention, an operator can more easily select an efficient procedure for having a robot perform a specific task than before.

1 is a diagram showing an example of the overall configuration of a motion plan optimization system 1. FIG. FIG. 3 is a diagram showing an example of a mechanical model and a control model of a robot. FIG. 2 is a diagram showing an example of the hardware configuration of a computer. FIG. 2 is a diagram showing an example of a functional configuration of a computer. FIG. 3 is a diagram showing an example of movement of a first workpiece and a second workpiece. It is a figure showing an example of a work log. FIG. 2 is a diagram showing an example of a Petri net. FIG. 2 is a diagram showing an example of a Petri net having multiple branches. 3 is a flowchart illustrating an example of the overall processing flow according to a motion planning program. FIG. 3 is a diagram showing an example of a firing sequence.

[1. Overall system configuration]
FIG. 1 is a diagram showing an example of the overall configuration of a motion plan optimization system 1. As shown in FIG. FIG. 2 is a diagram showing an example of a mechanism model and a control model of the robot 3. FIG. 3 is a diagram showing an example of the hardware configuration of the computer 2. As shown in FIG. FIG. 4 is a diagram showing an example of the functional configuration of the computer 2. As shown in FIG.

A motion plan optimization system 1 is a system for optimizing a robot's motion plan, and as shown in FIG. 1, it is composed of a computer 2 and a robot 3.

The robot 3 is a robot for which an optimal motion plan is calculated. Hereinafter, a case will be described using as an example a case where the robot 3 is a 6-axis robot installed on a horizontal plane.

As shown in FIG. 1 or 2, the robot 3 includes a base 30, a first arm 311, a second arm 312, a third arm 313, a fourth arm 314, a fifth arm 315, a tool 32, First drive unit 331, second drive unit 332, third drive unit 333, fourth drive unit 334, fifth drive unit 335, sixth drive unit 336, controller 34, communication interface 35, etc. Consisted of. Note that in FIG. 2, dotted lines represent wired or wireless communication paths.

In this embodiment, the base 30 is installed on a horizontal surface. The first arm 311 has its base end supported by the base 30 and rotates about a first axis 391 that is perpendicular to the installation surface (horizontal surface) of the base 30 as a rotation axis. The first drive unit 331 includes a motor, a speed reducer, an angle sensor, and the like, and rotates the first arm 311. Note that, like the first drive unit 331, the second drive unit 332 to the sixth drive unit 336 are also configured by a motor, a speed reducer, an angle sensor, and the like.

The second arm 312 has a proximal end supported by the distal end of the first arm 311, and a second axis 392 that is perpendicular to both the longitudinal direction of the first axis 391 and the second arm 312. Rotate using the axis as the axis. The second drive unit 332 pivots the second arm 312.

The third arm 313 has its base end supported by the distal end of the second arm 312, and pivots about a third axis 393 parallel to the second axis 392 as a swing axis. The third drive unit 333 pivots the third arm 313.

The fourth arm 314 has its base end supported by the distal end of the third arm 313 and rotates about a fourth axis 394 parallel to the longitudinal direction of the third arm 313 as a rotation axis. Fourth drive unit 334 rotates fourth arm 314 .

The fifth arm 315 has its base end supported by the distal end of the fourth arm 314 and pivots about a fifth axis 395 perpendicular to the fourth axis 394 as a swing axis. The fifth drive unit 335 pivots the fifth arm 315.

The tool 32 has its base end supported by the distal end of the fifth arm 315 and rotates about a sixth axis 396 parallel to the longitudinal direction of the fifth arm 315 as a rotation axis. A sixth drive unit 336 rotates the tool 32. The tool 32 also includes a hand consisting of a plurality of fingers and a motor that opens and closes the hand, and picks up and releases objects.

The angle sensor of the first drive unit 331 detects the angle θ ₁ between the reference posture and the current posture of the first arm 311 in the coordinate system of the base 30 . In other words, it is detected how much the first arm 311 has rotated from the reference posture. The angle sensor of the second drive unit 332 detects the angle θ ₂ between the first arm 311 and the second arm 312.

The angle sensor of the third drive unit 333 detects the angle θ ₃ formed by the second arm 312 and the third arm 313. The angle sensor of the fourth drive unit 334 detects the angle θ ₄ between the reference attitude of the fourth arm 314 in the coordinate system of the third arm 313 and the current attitude. In other words, it is detected how much the fourth arm 314 has rotated from the reference posture.

The angle sensor of the fifth drive unit 335 detects the angle θ ₅ formed by the fourth arm 314 and the fifth arm 315. The angle sensor of the sixth drive unit 336 detects the angle θ ₆ between the reference attitude of the tool 32 in the coordinate system of the fifth arm 315 and the current attitude. In other words, it is detected how much the tool 32 has rotated from the reference posture.

The angle sensors of each of the first drive unit 331 to the sixth drive unit 336 transmit the detected angles θ ₁ to θ ₆ to the computer 2 via the communication interface 35. The postures of the robot 3 include the posture of the first arm 311 with respect to the base 30, the posture of the second arm 312 with respect to the first arm 311, the posture of the third arm 313 with respect to the second arm 312, and the posture of the third arm 313 with respect to the second arm 312. The posture of the fourth arm 314 relative to the fourth arm 314 , the posture of the fifth arm 315 relative to the fourth arm 314 , and the posture of the tool 32 relative to the fifth arm 315 . Therefore, the posture of the robot 3 is specified by the angles θ ₁ to θ ₆ .

The controller 34 controls the first drive unit 331 to the sixth drive unit 336 so that the first arm 311 to the fifth arm 315 and the tool 32 take the postures as instructed by the computer 2. It also controls the motor of the tool 32 so that the hand of the tool 32 opens and closes according to instructions from the computer 2.

The communication interface 35 is a wired standard interface device such as a USB (Universal Serial Bus) adapter or NIC (Network Interface Card), or a wireless standard communication device such as a Bluetooth adapter or a Wi-Fi adapter, and is connected to the computer 2. Send and receive data between.

The computer 2 calculates an optimal motion plan for the robot 3 by process mining. An example in which a laptop personal computer is used as the computer 2 will be described below.

As shown in FIG. 3, the computer 2 includes a main processor 20, a RAM (Random Access Memory) 21, a ROM (Read Only Memory) 22, an auxiliary storage device 23, a communication interface 24, a display 25, a keyboard 26, and a pointing device 27. It is composed of etc.

In addition to the operating system, computer programs such as a motion planning program 40 are installed in the ROM 22 or the auxiliary storage device 23.

RAM 21 is the main memory of the computer 2. Computer programs such as the motion planning program 40 are loaded into the RAM 21 as appropriate.

Main processor 20 executes a computer program loaded into RAM 21. As the main processor 20, a GPU (Graphics Processing Unit), a CPU (Central Processing Unit), or the like is used.

The communication interface 24 sends and receives data to and from the robot 3. As the communication interface 24, a communication device of the standard adopted by the robot 3 is used.

The display 25 displays a screen for inputting commands or data, a screen showing the results of calculations by the main processor 20, or the like.

The keyboard 26 and pointing device 27 are input devices for an operator to input commands, data, and the like.

According to the motion planning program 40, the work log acquisition section 401, the work log storage section 402, the identifier assignment section 403, the event log generation section 404, the Petri net generation section 405, the Petri net storage section 406, and the optimal motion plan determination shown in FIG. Functions such as a section 407, an operation verification section 408, and an operation program generation section 409 are realized. Then, the following five services are provided, and the motion plan of the robot 3 is optimized.
- Acquisition of event log of robot 3 - Generation of Petri net - Determination of optimal motion plan by Petri net simulation - Generation of motion program - Implementation of motion program in robot 3 Below, work log acquisition unit 401 to motion program generation unit 409 Each function and each of the five services will be explained in turn.

[2. Processing of each part]
[2.1 Obtaining event log of robot 3]

(1) Collection of raw data FIG. 5 is a diagram showing an example of movement of the first workpiece 51 and the second workpiece 52. FIG. 6 is a diagram showing an example of work logs 61 and 62.

The work log acquisition unit 401 acquires work logs when the robot 3 executes a specific work in various procedures. Hereinafter, collection of work logs will be described using as an example the work of moving the first workpiece 51 and the second workpiece 52 from the first position and the second position to the third position, respectively.

Various procedures are possible to accomplish this task. There are multiple ways to move the first workpiece 51 and the second workpiece 52. For example, the first workpiece 51 and the second workpiece 52 can be moved as shown in FIG. 5(A) or FIG. 5(B). When moving the first workpiece 51 to the third position and then moving the second workpiece 52 to the third position as shown in FIG. 5(A), the robot 3 operates as follows, for example. do.
#01: Take the initial posture #02: Take the posture of picking up the first workpiece 51 with the hand of the tool 32 #03: Close the hand and pick the first workpiece 51 #04: The first workpiece 51 Take a position to relay the work 51 from the first position to the second position #05: Take a position to place (place) the first work 51 onto the second work 52 #06: Open the hand and take the position Place the first workpiece 51 onto the second workpiece 52 #07: Take a posture to pick the first workpiece 51 and the second workpiece 52 #08: Pick both workpieces #09: Place both workpieces on the second workpiece 52 Take a posture to relay from the position to the third position #10: Take a posture to place both workpieces to the third position #11: Open the hand and place both workpieces to the third position #12: Movement take the finishing position

In this embodiment, it is assumed that the hand of the tool 32 is in an open state in the initial posture. The "relaying posture" in steps #04 and #09 is the posture in which the tool 32 passes through the relaying position when moving the workpiece from the source (position before relaying) to the destination (position after relaying). be. The relay position may be the exact middle position on the straight line connecting the movement source and movement destination, or if there is an obstacle between the movement source and movement destination, It may be located in the sky. The same applies to steps #24, #27, and #30, which will be described later.

Alternatively, when placing the first workpiece 51 on top of the second workpiece 52 and moving both workpieces together to the third position as shown in FIG. 5(B), the robot 3 may, for example, works.
#21: Take the initial posture #22: Take the posture of picking the first work 51 with the hand of the tool 32 #23: Close the hand and pick the first work 51 #24: Pick the first work 51 Take a position to relay from the first position to the third position #25: Take a position to place the first workpiece 51 to the third position #26: Open the hand and move the first workpiece 51 to the third position Place to position #27: Take a posture to move the tool 32 from the third position to the second position (relay posture) #28: Take a posture to pick the second workpiece 52 with the hand #29: Take the hand Close and pick the second work 52 #30: Take a posture to relay the second work 52 from the second position to the third position #31: Place the second work 52 to the third position Take a posture #32: Open the hand and place the second workpiece 52 to the third position #33: Take a posture to end the movement

By the way, the first workpiece 51 and the second workpiece 52 can also be moved by procedures other than steps #01 to #12 and steps #21 to #33. For example, the second work 52 may be moved before the first work 51. Alternatively, the first workpiece 51 and the second workpiece 52 may be temporarily placed at the fourth position, and both workpieces may be moved together from the fourth position to the third position. There may also be multiple ways to avoid obstacles. For example, the tool 32 may take a posture in which it straddles an obstacle, or may take a posture in which it detours around the obstacle.

Therefore, the procedure for moving the two workpieces to a predetermined position (third position) is not limited to the above-mentioned steps #01 to #12 and steps #21 to #33. It can also be a street. However, in the present embodiment, for the sake of simplicity, a case will be described using as an example a case where work logs are obtained for two procedures: steps #01 to #12 and steps #21 to #33.

The operator teaches the robot 3 how to operate the robot 3 in each of steps #01 to #12, and the work log acquisition unit 401 acquires information obtained from the robot 3 at that time and stores it in the work log storage unit as a work log. 402. Similarly, the operator teaches the robot 3 how to operate the robot 3 in each of steps #21 to #33, and the work log acquisition unit 401 acquires the information obtained from the robot 3 at that time and records the work as a work log. It is stored in the log storage unit 402. The teaching method is the same as in conventional online teaching, and the operator can teach by touching and moving each part of the robot 3 directly (that is, manually), or by giving commands from the operating terminal. It may also be taught by

For example, the operator instructs each step of the work according to steps #01 to #12 as follows, and the work log acquisition unit 401 acquires the work log as follows.

The operator causes the robot 3 to assume an initial posture in order to teach the operation in step #01. Then, "pose" is input as the action type to the computer 2 or the operation terminal.

Then, the work log acquisition unit 401 causes the robot 3 to detect and acquire the angles θ ₁ to θ ₆ at the time when the robot 3 takes the initial posture. Then, motion data is generated that indicates the time when the initial posture is taken as the completion time, indicates the acquired angles θ ₁ to θ ₆ as parameters, and indicates the input motion type, and stores it in the work log storage unit 402. (See FIG. 6(A)).

The operator makes the robot 3 take a posture in which it can pick the first workpiece 51 with the tool 32 in order to teach the operation in step #02. Then, input "pose" as the motion type.

Then, the work log acquisition unit 401 causes the robot 3 to detect and acquire the angles θ ₁ to θ ₆ at the time when the robot 3 takes this posture, as in step #10. Then, motion data is generated that indicates the time at this point as the completion time, indicates the acquired angles θ ₁ to θ ₆ as parameters, and indicates the input motion type, and stores it in the work log storage unit 402.

The operator closes the hand of the tool 32 to pick the first workpiece 51 in order to teach the operation of step #03. Then, input "grasp" as the operation type.

Then, the work log acquisition unit 401 generates action data that indicates the time when the hand is closed as the completion time and the input action type, and stores it in the work log storage unit 402. Note that a value indicating the degree of closing of the hand of the tool 32 may be shown as a parameter in the motion data. The same applies hereafter.

In order to teach the operation in step #04, the operator causes the robot 3 to assume a posture in which the tool 32 is placed at a relay position to the second position. Then, input "pose" as the motion type.

Then, similarly to steps #01 and #02, the work log acquisition unit 401 causes the robot 3 to detect and acquire the angles θ ₁ to θ ₆ at the time when the robot 3 takes this posture. Then, motion data is generated that indicates the time at this point as the completion time, indicates the acquired angles θ ₁ to θ ₆ as parameters, and indicates the input motion type, and stores it in the work log storage unit 402.

The operator makes the robot 3 assume a posture in which it can place the first workpiece 51 on the second workpiece 52 in order to teach the operation in step #05. Then, input "pose" as the motion type.

Then, similarly to steps #01, #02, and #04, the work log acquisition unit 401 causes the robot 3 to detect and acquire the angles θ ₁ to θ ₆ at the time when the robot 3 takes this posture. Then, motion data is generated that indicates the time at this point as the completion time, indicates the acquired angles θ ₁ to θ ₆ as parameters, and indicates the input motion type, and stores it in the work log storage unit 402.

The operator opens the hand of the tool 32 to place the first workpiece 51 in order to teach the operation in step #06. Then, input "release" as the operation type.

Then, the work log acquisition unit 401 generates action data that indicates the time when the hand is opened as the completion time and the input action type, and stores it in the work log storage unit 402.

Instead of the operator inputting the motion type, the work log acquisition unit 401 may determine the motion type based on the movement of the robot 3. For example, when the operator stops moving the robot 3, it is determined that it is in "pose". Alternatively, if the operator moves the hand of the tool 32 in the closing direction, it is determined as "grasp". Alternatively, if the operator moves the hand of the tool 32 in the direction of opening, it is determined as "release".

For steps #06 and subsequent steps, the operator instructs the robot 3 in the same manner, and the work log acquisition unit 401 generates movement data and stores it in the work log storage unit 402. As a result, 12 pieces of operation data as shown in FIG. 6(A) are stored in the work log storage unit 402. A collection of these operation data is the work log 61 related to the procedure of steps #01 to #12. Note that the "identifier" is added later to generate the Petri net. At the time the motion data is generated, the identifier is Null.

Further, the operator instructs the robot 3 in the same manner regarding the work performed in steps #21 to #33, and the work log acquisition unit 401 generates operation data and stores it in the work log storage unit 402. As a result, 13 pieces of operation data as shown in FIG. 6(B) are stored in the work log storage unit 402. A collection of these operation data is the work log 62 related to the procedure of steps #21 to #33.

Note that instead of moving each part of the robot 3, the operator may teach the work by inputting motions (posture, opening/closing, etc.) to the simulator of the robot 3. Then, the work log acquisition unit 401 may generate motion data by determining the angles θ ₁ to θ ₆ or the motion type based on the motion input to the simulator. Furthermore, the motion of the robot 3 may be simulated on a simulator as input, and the time at which the motion of each step is completed may be indicated in the motion data as the completion time.

(2) Data processing The identifier assigning unit 403 and the event log generating unit 404 process the data stored in the work log storage unit 402 in order to generate a Petri net. The processing of the identifier assigning section 403 and the event log generating section 404 will be described below, taking as an example the case where the work logs 61 and 62 shown in FIG. 6 are processed.

The identifier assigning unit 403 assigns a unique identifier to each action represented by each action data included in the work log 61 and each action represented by each action data included in the work log 62. However, the same identifier is assigned to the same operation. For example, assign an identifier as follows.

The identifier assigning unit 403 assigns a unique identifier to each action represented by each action data included in the work log 61. That is, since the work log 61 includes 12 pieces of action data, "a", "b", "c", ..., "j", "k", "l" are written for each of the 12 actions. ” is assigned an identifier.

Further, the identifier assigning unit 403 selects, from among the actions represented by each of the action data included in the work log 62, the action that is the same as the action represented by any of the action data included in the work log 61. Specifically, those having the same motion type and the same posture during motion are selected. Then, an identifier of the selected action is assigned.

For example, in the operation at completion time T ₂₁ (step #21), the operation type is "pose" and the angles (θ ₁ , θ ₂ , θ ₃ , θ ₄ , θ ₅ , θ ₆ ) are (θ _{1_21} , θ _{2_21} , θ _{3_21} , θ _{4_21} , θ _{5_21} , θ _{6_21} ). On the other hand, for the operation at completion time T ₀₁ (step #01), the operation type is "pose" and the angles (θ ₁ , θ ₂ , θ ₃ , θ ₄ , θ ₅ , θ ₆ ) are (θ _{1_01} , θ _{2_01} , θ _{3_01} , θ _{4_01} , θ _{5_01} , θ _{6_01} ).

That is, the operation type at completion time _T21 is the same as the operation type at completion time _T01 . Therefore, if the posture at completion time _T21 is the same as the posture at completion time _T01 , the motion at completion time _T21 is selected as the same motion as the motion at completion time _T01 .

However, when the operator moves the robot 3 to obtain a work log, a slight deviation may occur even if the operator intends to have the robot 3 assume the same posture at both completion times.

Therefore, the identifier assigning unit 403 includes the absolute value of the difference between angle θ _{1_01} and angle θ _{1_21} , the absolute value of the difference between angle θ _{2_01} and angle θ _{2_21} , the absolute value of the difference between angle θ _{3_01} and angle θ _{3_21} , The absolute value of the difference between angle θ _{4_01} and angle θ _{4_21} , the absolute value of the difference between angle θ _{5_01} and angle θ _{5_21} , and the absolute value of the difference between angle θ _{6_01} and angle θ _{6_21} are all the same. If the value is within an allowable range, the posture at the completion time _T21 and the posture at the completion time _T01 are considered to be the same. In the following, a case will be explained in which both postures are considered to be the same. Therefore, the identifier assigning unit 403 selects the operation at the completion time T ₀₁ as the same operation as the operation at the completion time T ₂₁ . Then, the identifier of the operation at the completion time T ₀₁ is assigned to the operation at the completion time T ₂₁ . In this example, "a" is given.

The same method is used for each operation after the completion time _T22 , by comparing the operation type and posture of each operation shown in the work log 61, selecting the same one, and assigning an identifier. However, the posture of the motion whose motion type is "grasp" or "release" is represented by the parameter of the motion data of the most recent motion whose motion type is "pose." For example, the posture of the motion at the completion time _T03 is represented by the parameters of the motion data at the completion time _T02 .

Hereinafter, the same operations as those at the completion times T ₂₂ , T ₂₃ , T ₂₈ , T ₂₉ , T ₃₀ , T ₃₁ , T ₃₂ , and T ₃₃ are performed at the completion times T ₀₂ , T ₀₃ , T ₀₇ , An example will be described in which the operations T ₀₈ , T ₀₉ , T ₁₀ , T ₁₁ , and T ₁₂ are selected and assigned identifiers as shown in FIG. 6(B).

Further, the identifier assigning unit 403 assigns a unique identifier to an action that is not the same as any action indicated in the action data included in the work log 61 among the actions indicated in the action data included in the work log 62. do. Hereinafter, a case will be described using as an example a case where each operation at completion times T ₂₄ to T ₂₇ is not the same as any operation at completion times T ₀₁ to T ₁₂ . Therefore, the identifier assigning unit 403 assigns a unique identifier to each operation from completion time T ₂₄ to T ₂₇ , as shown in FIG. 6(B).

By the way, as described later, the α algorithm is used in this embodiment. In the α algorithm, each step corresponds to a "task". Therefore, hereinafter, the operation of each step may be referred to as a "task".

The event log generation unit 404 generates an event log L based on the work logs 61 and 62 and the identifier assigned by the identifier assignment unit 403. Event log L is
{(a,b,c,d,e,f,g,h,i,j,k,l),(a,b,c,n,o,p,m,g,h,i,j, k,l)}
represents. Of these, (a, b, c, d, e, f, g, h, i, j, k, l) is the time identifier of each task (action) shown in the action data included in the work log 61. They are arranged in order. In addition, (a, b, c, n, o, p, m, g, h, i, j, k, l) is the identifier of each task shown in the operation data included in the work log 62 arranged in chronological order. It is something that

The serial path in event log L is A={a,b,c} B={d,e,f} C={n,o,p,m} D={g,h,i,j,k ,l}, the event log L can be expressed as {(A,B,D),(A,C,D)}.

[2.2 Generation of Petri net]
FIG. 7 is a diagram showing an example of the Petri net 7.

The Petri net generation unit 405 generates a Petri net 7 as shown in FIG. 7 by outputting model components from the order relationship between tasks in the event log L based on the α algorithm. Petri net 7 is stored in Petri net storage section 406.

The α algorithm is a well-known process mining algorithm for generating a Petri net (Petri net model) as a process model, and is explained in the following known documents 1 and 2. Furthermore, an example of a method for implementing the α algorithm in a computer program is described in Publication No. 3.
[Known Document 1] W. van der Aalst, T. Weijters, L. Maruster, “Workflow mining: discovering process models from event logs”, IEEE Transactions on Knowledge and Data Engineering, Vol. 16, No. 9, pp. 1128 -1142, (2004).
[Known Document 2] Tadashi Iijima, Keiichi Tabata, Shinobu Saito, “Process Mining Survey (No. 4: Algorithm (1))”, Journal of the Society of Information Systems, Vol. 13, No. 1, pp. 43-45, (2017). https://www.issj.net/journal/jissj/Vol13_No1_Open/A4V13N1.pdf
[Public Publication 3] "Implementing and understanding the process mining alpha algorithm in Python", https://ownsearch-and-study.xyz/2019/11/06/python-processmining-alpha-implementation/
Petri Net 7 is a mathematical model for representing concurrent, asynchronous, distributed, parallel, deterministic, and stochastic systems, and consists of two types of nodes: places representing conditions and transitions representing events. is a bipartite directed graph with Alternatively, it can be said to be a model of a discrete event system. In FIG. 7, circular nodes are places and rectangular nodes are transitions. Places and transitions are connected by arcs. Arcs are represented by line segments with arrows. In this way, the structure of the Petri net 7 is specified by places, transitions, and arcs.

The black dots in the places are tokens. There are multiple (two) arcs extending from the place p4, but this represents that the tokens selectively advance only to one, rather than to multiple tokens. Although not shown in FIG. 7, if a plurality of arcs extend from the transition, there will be a plurality of tokens, and the tokens will advance one by one to each destination place.

The Petri net generation unit 405 calculates sets of places, transitions, and arcs P _L , T _L , and F _L from the event log L based on the α algorithm. In this example, P _L , T _L , and F _L are P _L =[[A,{B,C}],[{B,C},D],i _L ,o _L ]
T _L = {A, B, C, D}
F _L = [A[A,{B,C}],[[A,{B,C}]B],[[A,{B,C}]C],[B[{B,C}, D]],
[C[{B,C},D], [[{B,C},D],D],[i _L ,A],[D,o _L ]]]
It is calculated as follows. The Petri net 7 can be drawn as shown in FIG. 7 based on P _L , T _L , and F _L . In addition, in the Petri net simulation described below, the Petri net 7 is represented by a connection matrix of M _T ×M _P and used. “M _T ” and “M _P ” are the numbers of transitions and places of the Petri net 7, respectively, and in the example of FIG. 7, both are “16”.

The Petri net 7 is converted into an image and displayed on the display 25 as shown in FIG. Or printed by a printer.

[2.3 Determination of optimal motion plan using Petri net simulation]
The optimal motion plan determining unit 407 determines the optimal motion plan for the work to be performed by the robot 3. In the following, the optimal motion plan determination will be described using an example of determining the optimal motion plan for moving the first workpiece 51 and the second workpiece 52 from the first position and the second position to the third position, respectively. The processing of section 407 will be explained.

The optimal motion plan determining unit 407 randomly selects N firing sequences in which the positions of the initial token and the target token are p1 and p16 from the Petri net 7, respectively.

Specifically, the optimal motion plan determining unit 407 places a token at the place p1 and transitions the token to the place p16 according to the firing rule of the Petri net. On the way, each time the token reaches a place that branches into multiple arcs, one of these arcs is randomly selected and the token is advanced to the transition to which the selected arc is connected. One firing sequence can be obtained by performing such a transition once. By repeating such transitions, it is possible to obtain and select N firing sequences. A firing sequence can be said to be a path from an initial place to a target place. Note that N is an integer of 2 or more. If the specifications of the computer 2 or the processing time are sufficient, all firing sequences may be acquired and selected.

In this embodiment, as shown in FIG. 7, there is only one place where the arc branches, the place p4, and the number of branches is two. Therefore, at most two firing sequences are selected. Hereinafter, a case where two firing sequences, the {A, B, D} firing sequence and the {A, C, D} firing sequence, are selected will be described as an example.

For each selected firing sequence, the optimal motion plan determination unit 407 verifies the token transition according to the firing rule of the Petri net, and calculates the time required from when the token is placed at p1 to when it reaches p16 (hereinafter referred to as "total time"). Calculate the firing transition time.

In order to calculate the total firing transition time, it is necessary to know the firing transition time of each transition in the firing sequence. Therefore, in this embodiment, the firing transition time of each transition is determined as follows.

The firing transition time of the transition (the "grasp" task) corresponding to the action of picking up the workpiece is a predetermined time _Ta . Similarly, the firing transition time of the transition (task "release") corresponding to the action of placing the workpiece is the predetermined time _Tb . The times T _a and T _b may be the same length, such as "1 second", or may be different lengths.

Furthermore, the firing transition time of the transition (the "pose" task) corresponding to the action of taking a posture specified by angles θ ₁ to θ ₆ is the time when that task is completed and the time when the task immediately before that task is completed. This is the difference between For example, the firing transition time of the transition "b" is the difference between the completion time T ₀₂ of the task of the transition "b" (see FIG. 6) and the completion time T ₀₁ of the task of the transition immediately before it, that is, the transition "a". The difference is "T ₀₂ - _{T 01} ". However, the firing transition time of the first transition is the difference between the task time and the work start time of that transition. For example, if the start time of the work in steps #01 to #12 is T ₀₀ , the firing transition time of the transition “a” is “T ₀₁ −T ₀₀ ”.

There are some transitions for which the firing transition time can be calculated based on either the work log 61 or the work log 62, but in such a case, it may be based on either. For example, the firing transition time of the transition “b” may be “T ₀₂ −T ₀₁ ” or “T ₂₂ −T ₂₁ ”. However, it is desirable to use one of them in unison.

Then, the optimal motion plan determination unit 407 selects the firing sequence with the shortest total firing transition time as the optimal firing sequence, and determines (estimates) that the motion procedure specified by the optimal firing sequence is the optimal motion plan for the task. )do.

The optimal motion plan determining unit 407 calculates (a, b, c, d, e, f, g, h, i, j, k, l) as the total firing transition time of the firing sequence of {A, B, D}. Calculate the total firing transition time of each transition, and calculate the total firing transition time of the firing sequence {A, C, D} as (a, b, c, n, o, p, m, g, h, i, j ,k,l), the total firing transition time of each transition is calculated. Then, the one with the shorter total firing transition time is selected as the optimal firing sequence.

In this embodiment, both the firing sequence of {A, B, D} and the firing sequence of {A, C, D} are the steps in which the operator moves the robot 3 to obtain the work log 61 or the work log 62. Therefore, the optimal firing sequence also matches either procedure.

FIG. 8 is a diagram showing an example of a Petri net 71 having multiple branches.

However, for example, the work log of (a ₀ ,b ₁ ,c ₀ ,d ₁ ,e ₀ ,f ₁ ,g ₀ ,h ₁ ,i ₀ ,j ₁ ,k ₀ ) and (a ₀ ,b ₂ ,c ₀ , _d2 , _e0 , f2, _g0 , _{h2 , i0} , _j2 , _k0 ), a Petri net 71 as shown in FIG. 8 is obtained. Since the Petri net 71 includes three branches that divide into two, there are 2 to the third power, that is, eight firing sequences. Therefore, a firing sequence that does not match either work log may be the optimal firing sequence. As the number of work logs included in the event log increases or as the number of branches within the work log increases, there is a tendency for the firing sequence pattern to increase exponentially.

[2.4 Generation of operation program]
The motion verification section 408 is a simulator that verifies the consistency of the model, and verifies by simulation whether the robot 3 can complete the work according to the optimal firing sequence selected by the optimal motion plan determination section 407. For example, if the firing sequence {A, C, D} is selected as the optimal firing sequence, (a, b, c, n, o, p, m, g, h, i, j, k, l) It is verified whether the robot 3 can complete the work according to the specifications by performing the tasks (movements) corresponding to each transition.

In particular, as shown in FIG. 8, when the number of firing sequences that can be selected from the Petri net is greater than the number of work logs, the optimal firing sequence may not match the procedure of any of the work logs. Therefore, it is important for the operation verification unit 408 to simulate and verify the work performed by the robot 3 based on the optimal firing sequence. Furthermore, as will be described later, the Petri net 7 may be applied to other tasks or to other robots. In such cases, it is important to simulate and verify the work.

When the operation verification unit 408 verifies that the robot 3 can perform the work according to the specifications based on the optimal firing sequence, the operation program generation unit 409 generates an operation program 80 for operating the robot 3 according to the optimal firing sequence. generate. Note that the operation program 80 is tuned so that the robot 3 can move the first workpiece 51 and the second workpiece 52 more reliably.

[2.5 Implementation of motion program to robot 3]
The computer 2 can cause the robot 3 to perform work by controlling the robot 3 based on the operation program 80.

[3. Overall processing flow and effects of this embodiment]
FIG. 9 is a flowchart illustrating an example of the overall processing flow by the motion planning program 40.

Next, the overall processing flow of the computer 2 by the motion planning program 40 will be explained with reference to a flowchart. The computer 2 executes processing according to the procedure shown in FIG. 9 based on the motion planning program 40.

The operator teaches the robot 3 a plurality of work procedures for a specific purpose. The computer 2 acquires data of each procedure as a work log (#101 in FIG. 9) and generates an event log (#102). Then, a Petri net is generated based on the event log (#103).

Furthermore, the computer 2 selects N firing sequences based on the Petri net, calculates the total firing transition time of each firing sequence, and selects the firing sequence with the shortest total firing transition time as the optimal firing sequence (# 104). Operation verification is performed by simulating the work of the robot 3 according to the optimal firing sequence (#105).

If the operation verification is successful, the computer 2 generates an operation program 80 for operating the robot 3 according to the optimal firing sequence (#106), and applies it to the robot 3 (#107). The robot 3 is controlled based on an operation program 80 and performs work for a specific purpose.

According to this embodiment, the Petri net 7 for a specific task is generated by the motion plan optimization system 1 based on the task logs 61 and 62. As illustrated in FIG. 7, the Petri net 7 can visually represent, as one graph, a plurality of firing sequences (paths) that can be taken to complete a specific task. Therefore, by referring to the Petri net 7, the operator can select an efficient procedure more easily than before.

Furthermore, an optimal firing sequence is selected from the Petri net 7, its feasibility on the robot 3 is verified, and it is applied to the robot 3. Therefore, the operator can more easily select an efficient procedure from among a plurality of firing sequences depending on the situation.

[4. Modifications and application examples]
FIG. 10 is a diagram showing an example of the firing sequence 72.

In this embodiment, as shown in FIG. 7, a Petri net in which only one token is used has been described as an example, but the present invention can also be applied to a Petri net in which a plurality of tokens are used. . In this case as well, the optimal action plan determining unit 407 (see FIG. 4) may calculate the time required for the token to transition from the initial place of the firing sequence to the target place as the total firing transition time according to the firing rule. For example, if the firing sequence 72 shown in FIG. 10 is selected from the Petri net, the total firing transition time may be calculated as follows.

In firing sequence 72, the tokens become two at transition t ₁ and advance to places p ₂ and p ₄ , respectively. Then, merge at place _p6 . The optimal motion plan determination unit 407 determines the firing transition time T ₁ of the transition t ₁ , the firing transition time T ₂ of the transition t ₂ , and the firing transition time T ₃ of the transition t ₃ , whichever is longer, and the firing transition time T 1 of the transition t 1 . ₄ and the firing transition time _T4 is calculated as the total firing transition time of the firing sequence 72. That is, the longer one of T ₁ +T ₂ +T ₄ and T ₁ +T ₃ +T ₄ is calculated as the total firing transition time.

For example, transition _t2 corresponds to the task of taking a posture to pick the workpiece with the hand of the tool 32, and transition _t3 corresponds to the task of photographing the workpiece with the camera of the tool 32 to confirm that there are no defects in the workpiece. do.

The Petri net 7 was generated to optimize the motion plan for moving the first workpiece 51 and the second workpiece 52 from the first position and the second position to the third position, respectively. Petri net 7 can be applied for other tasks. For example, it may be used to optimize a motion plan for moving the first workpiece 51 and the second workpiece 52 from the fourth position and the fifth position to the sixth position, respectively.

In this case, the parameters of the work logs 61 and 62 (see FIG. 6) are changed in accordance with the posture that the robot 3 takes at each position. Furthermore, the operation time of each step is changed. The posture and operation time may be determined by manually moving the robot 3, or may be determined using a simulator. Alternatively, it may be obtained by substituting it into a conversion formula. The operation planning program 40 may be configured to implement a work log change means (not shown), and parameters and operation times may be changed by the work log change means.

Then, the optimal action plan determining unit 407 calculates the total firing transition time of each of the N firing sequences based on the changed parameters and operation time, and selects the optimal firing sequence.

Alternatively, the Petri net 7 may be used for optimal motion planning when a robot other than the robot 3 performs a similar task. In this case as well, the parameters of the work logs 61 and 62 (see FIG. 6) may be changed in accordance with the posture taken by the robot at each position, and the operation time of each step may also be changed. Then, the optimal operation plan determining unit 407 calculates the total firing transition time of each of the N firing sequences based on the changed parameters and operation time, and selects the optimal firing sequence.

In this way, the Petri net 7 can be extended for other tasks on the robot 3 or for tasks on other robots.

Although the present embodiment has been described as an example in which the Petri net 7 is generated based on two work logs 61 and 62, it may be generated based on three or more work logs.

In this embodiment, an example is explained in which a motion plan is optimized to have the robot 3 perform a task of moving two workpieces located at different positions. The present invention can also be applied when optimizing a motion plan.

In the present embodiment, the optimal action plan determining unit 407 selects the firing sequence with the shortest total firing transition time as the optimal firing sequence, but selects a plurality of firing sequences whose total firing transition time falls within the requested time. The firing sequence may be presented to the operator and the operator may select the optimal firing sequence from among these firing sequences.

The data of the Petri net 7 may be output from the computer 2 to another computer, and the other computer may use the Petri net 7 to optimize the robot's motion plan.

In this embodiment, the case where the robot 3 is a 6-axis robot has been explained as an example, but if the robot 3 is an articulated robot with 7 or more axes, a dual-arm robot, a SCARA robot, or a parallel link robot. The present invention can also be applied in this case.
In this embodiment, the firing sequence with the shortest total firing transition time, that is, the required time, is selected as the optimal firing sequence, but it may be selected based on other indicators. For example, the firing sequence that requires the least amount of power to change the posture or the firing sequence that requires the shortest movement distance of the tool 32 may be selected as the optimal firing sequence. Alternatively, a comprehensive movement cost may be calculated based on the total firing transition time, power consumption, moving distance, etc., and the firing sequence with the lowest movement cost may be selected as the optimal firing sequence.

In addition, the configuration of the entire motion plan optimization system 1, the computer 2, and the robot 3 or each part, the content of processing, the order of processing, the data configuration, the method of determining the firing transition time, etc. may be changed as appropriate in accordance with the spirit of the present invention. can do.

1 Motion plan optimization system 3 Robot 401 Work log acquisition unit (procedural data acquisition means)
405 Petri net generation unit (process model generation means)
407 Optimal operation plan determining unit (eligible route selection means)
408 Operation verification section (verification means)
61 Work log (procedure data)
62 Work log (procedure data)
7 Petri net (process model)

Claims (9)

a procedure data acquisition means for acquiring procedure data indicating each of a plurality of procedures for causing a robot to perform a specific task;
process model generation means for generating a process model of the work based on the procedure data of each of the plurality of procedures;
A robot motion planning support system comprising: Qualified route selection means for selecting a route that satisfies predetermined requirements from all or part of a plurality of routes that can be taken from an initial node to a target node in the process model as a qualified route suitable for the work. ,
has,
The robot motion planning support system according to claim 1. The eligible route selection means randomly selects N routes from among the plurality of routes, and selects the route with the shortest required time or required cost for the robot among the N routes as the eligible route. elect,
The robot motion planning support system according to claim 2. The process model generation means generates a Petri net as the process model using an α algorithm,
The eligible route selection means selects the N routes by randomly advancing tokens in places having branches in the Petri net.
The robot motion planning support system according to claim 3. The procedure data acquisition means acquires, as the procedure data for each of the plurality of procedures, data indicating a completion time of each step constituting the procedure;
The eligible route selection means calculates the required time for each of the N routes based on the difference between the completion times of two consecutive steps among the steps.
The robot motion planning support system according to claim 3 or 4. verification means for verifying whether the work can be completed by performing a simulation of operating the robot according to the qualified route;
has,
A robot motion planning support system according to any one of claims 2 to 4. The qualified route selection means randomly selects N routes from among the plurality of routes as second qualified routes suitable for having a second robot different from the robot perform the work. , select the route with the shortest time or cost required by the second robot from among the N routes;
The robot motion planning support system according to claim 2. Providing procedural data to a computer that shows each of the multiple steps to have a robot perform a specific task,
causing the computer to execute a process of generating a process model of the work based on the procedure data of each of the plurality of procedures;
A robot motion planning support method characterized by: A computer program used in a computer that supports a motion plan for having a robot perform a specific task,
to the computer;
causing the robot to perform a process of acquiring procedure data indicating each of a plurality of procedures for causing the robot to perform the work;
executing a process of generating a process model of the work based on the procedure data of each of the plurality of procedures;
A computer program characterized by:

Images