WORKPIECE TRANSFER SYSTEM

Abstract

Workpiece transfer system (1), comprising: a workpiece feeding device (4) which is configured to feed at least one workpiece (71); a robot (2) which is configured to transfer the at least one workpiece (71) that is being fed in; a workpiece removal device (5) on which the at least one workpiece (71) transferred by the robot (2) from the workpiece feed device (4) is to be placed; a machine learning device (15) configured to construct a learning model used to specify an operating condition of the robot (2); and a control device (10) configured to select the operating condition of the robot (2) by applying a condition relating to the at least one workpiece (71) and / or a relationship relating to the robot (2) to the learning model, and to control the robot (2) according to the operating condition, wherein the machine learning device (15) includes an input data determination unit (16) configured to determine as input data a correspondence relationship between the condition relating to the at least one workpiece (71) and / or the condition relating to the robot (2) and the operating condition, a marker detection unit (17) configured to detect marker information indicating whether or not the processing power of the robot (2) is excessive or insufficient, and a learning unit (18) configured to conduct supervised learning based on teaching data constructed by a set of input data and labeling to construct the learning model to be used to specify the operating condition of the robot (2), and the control device (10) includes a storage unit (12) in which the learning model obtained from the machine learning device (15) is stored, a selection unit (111) that is configured to select an optimal operating condition for the robot based on the learning model, and an operating control unit (112) which is configured to control the operation of the robot according to the operating condition selected by the selection unit (111).

Description

[0001] The present invention relates to a workpiece transfer system.

[0002] In a case where the workpiece feed rate in a workpiece transfer system that includes a robot for picking up a workpiece can be higher or lower, the robot's operating condition is adjusted based on the highest workpiece feed rate.

[0003] However, if the workpiece feed rate is low, the robot's processing power becomes excessive relative to the workpiece feed rate. Consequently, the robot terminates the job early and enters a standby mode, remaining in this mode until the next workpiece is fed in. To enter standby mode, the robot must decelerate, and to exit standby mode, it must accelerate. This extra acceleration / deceleration leads to wasteful electrical consumption and causes unnecessary wear on the speed reducer. Increasing the robot's operating speed beyond what is necessary also results in wasteful electrical consumption and unnecessary wear on the speed reducer.

[0004] To cope with this, a known robot palletizer is configured to calculate a processing period of time per workpiece, to calculate and operate the robot at various speeds, such as an acceleration speed, a fixed speed and a deceleration speed, based on an operating pattern corresponding to the processing period of time (see, for example, patent document 1).

[0005] See also patent documents 2-7. Patent document 1: JP H06 - 126 668 A Patent document 2: DE 10 2018 215 057 A1 Patent document 3: DE 10 2018 007 107 A1 Patent document 4: DE 10 2017 008 836 A1 Patent document 5: DE 10 2016 013 731 A1 Patent document 6: DE 10 2015 002 571 B4 Patent document 7: EP 3 372 342 A1

[0006] However, according to the technology disclosed in the Japanese unexamined patent application, publication number JP H06-126668 A, each operating pattern is fixed according to the processing period per workpiece, and the modification or updating of the operating pattern and / or the operating condition according to the feed state of the workpiece is not taken into account.

[0007] Therefore, the task is to provide a workpiece transfer system that is able to flexibly change or update any operating pattern and / or operating condition based on the workpiece feed state and / or the robot's operating state.

[0008] This task is solved by a workpiece transfer system comprising: a workpiece feeder configured to feed at least one workpiece; a robot configured to transfer the fed workpiece; a workpiece discharge device on which the workpiece transferred by the robot from the workpiece feeder is to be placed; a machine learning device configured to construct a learning model used to specify an operating condition for the robot; and a control device configured to select the robot's operating condition by applying a condition relating to the workpiece and / or a relationship relating to the robot to the learning model and to control the robot according to the operating condition, wherein the machine learning device includes an input data acquisition unit.which is configured to determine, as input data, a correspondence relationship between the condition relating to at least one workpiece and / or the condition relating to the robot and the operating condition; a marker determination unit configured to determine, as marker information, whether or not the robot's processing performance is excessive or insufficient; and a learning unit configured to perform supervised learning based on training data built from a set of input data and the marker to construct the learning model to be used to specify the robot's operating condition; and the control device includes a storage unit in which the learning model determined from the machine learning device is stored; a selection unit configuredThe robot's operating condition is to be selected as an optimal operating condition based on the learning model, and an operating control unit is configured to control the robot's operation according to the operating condition selected by the selection unit.

[0009] According to one aspect of the present disclosure, it is possible to change or update each operating pattern and / or operating condition depending on the workpiece feeding state and / or the robot operating state. Fig. Figure 1 is a view showing an overall configuration of a workpiece transfer system according to one embodiment; Fig. 2 is a functional block diagram of a control device included in the workpiece transfer system according to the embodiment; Fig. 3 is a block diagram showing details of the control device and a machine learning device included in the workpiece transfer system according to the embodiment; Fig. Figure 4 is a view showing coordinate positions to be used to update an operating condition table that serves as a learning model in the workpiece transfer system according to the embodiment; Fig. 5A is a view showing a sequence of updates to the operating conditions table, which serves as the learning model in the workpiece transfer system, according to the embodiment; Fig. 5B is a view showing the process of updating the operating conditions table, which serves as the learning model in the workpiece transfer system, according to the embodiment; Fig. 5C is a view showing the process of updating the operating conditions table, which serves as the learning model in the workpiece transfer system, according to the embodiment; Fig. 6A shows an example of the operating conditions table, which serves as the learning model in the workpiece transfer system, according to the embodiment; Fig. Figure 6B shows an example of the operating conditions table, which serves as the learning model in the workpiece transfer system, according to the embodiment; Fig. Figure 7 is a view showing an overall configuration of a workpiece transfer system according to one embodiment; and Fig. Figure 8 is a view showing a pickup area in the workpiece transfer system according to the embodiment.

[0010] The following describes a first embodiment of the present invention with reference to Fig. 1 to 6B.

[0011] Fig. Figure 1 is a view showing an overall configuration of a workpiece transfer system 1 according to the first embodiment of the present invention.

[0012] As in Fig. As shown in Figure 1, the workpiece transfer system 1 according to the embodiment includes a robot 2, a detection device 3, a workpiece feed device 4, a workpiece discharge device 5, a control device 10 and a machine learning device 15.

[0013] The robot 2 picks up a workpiece 71 supplied by the workpiece feed device 4 and places the workpiece 71 on the workpiece discharge device 5.

[0014] The "pickup movement" for the workpiece 71 can be carried out specifically in the following manner, for example. That is, at a "pickup position," a vacuum condition is created between the workpiece 71 and a suction pad, which is provided at the distal end of a hand 23 (described later) and which is connected to a vacuum generating device via a polymer tube, and consequently the suction pad adheres to the workpiece 71 by suction in order to pick up the workpiece 71.

[0015] The "placing movement" for the workpiece 71 can, for example, be carried out specifically in the following manner. That is, at a "placing position," the vacuum between the suction pad and the workpiece 71 is released by switching off the vacuum generation device, and consequently, the workpiece 71 is placed.

[0016] The robot 2 comprises a robot body 21, an arm 22 and the hand 23. According to a command from the robot body 21, the robot 2 drives the arm 22 to perform the pick-up or place-down movement for the workpiece 71 with the hand 23, which is provided at the distal end of the arm 22.

[0017] The detection device 3 detects the workpiece 71 supplied by the workpiece feed device 4. More specifically, the detection device 3 detects the number of workpieces 71, the positions and orientation of the workpieces 71 and / or the like, and outputs the detection result to the control device 10.

[0018] The workpiece feeder 4 is a device for feeding the workpieces 71 to the workpiece discharger 5. For example, the workpiece feeder 4 can be constituted by a belt conveyor. In the description below, a workpiece to be picked up first by the robot 2 is referred to as a first workpiece 71a, and a workpiece to be picked up next by the robot 2 after the first workpiece is referred to as a second workpiece 71b. The first embodiment assumes that the workpieces 71 fed by the workpiece feeder 4 are arranged essentially in a row on the conveyor and that the conveyor operates at a higher speed than in a second embodiment (described later).

[0019] For the workpiece feeder 4, a zone 60 is defined where a workpiece can be picked up (hereinafter referred to as the workpiece picking zone 60). The workpiece picking zone 60 has a center point that coincides with a central axis of the workpiece discharge device 5, the central axis being orthogonal to the workpiece feeder 4. Additionally, the workpiece picking zone 60 is sandwiched between an upper boundary line 61 and a lower boundary line 62. The workpiece picking zone 60 includes a fault line 63, which indicates the lower limit below which the robot 2 will not fail to pick up the workpiece 71.

[0020] In the first embodiment, the time period required for a workpiece 71 to pass through the workpiece holding area 60 is assumed to be shorter than that of a second (later described) embodiment.

[0021] The workpiece discharge device 5 is a device for removing, outside the system, the workpiece 71 that has been picked up by the robot 2 from the workpiece feed device 4 and placed on the workpiece discharge device 5. For example, the workpiece discharge device 5 can be configured with a belt conveyor. In one example, a pallet 51 is placed on the workpiece discharge device 5, and the robot 2 picks up the workpiece 71 from the workpiece feed device 4 and places the workpiece 71 on the pallet 51.

[0022] The control device 10 is a device for controlling the entire workpiece transfer system 1. Fig. Figure 2 is a block diagram showing functions of the control device 10.

[0023] The control device 10 includes a control unit 11 and a storage unit 12.

[0024] The control unit 11 controls the entire control device 10 and implements various functions of the present embodiment by reading different programs from a memory area, such as a ROM, RAM, flash memory, or hard disk drive (HDD), and executing the programs as needed. The control unit can be a CPU. The control unit 11 includes a selection unit 111, an operating control unit 112, a monitoring unit 113, a change amount calculation unit 114, an electrical power amount calculation unit 115, and a feature amount output unit 116.

[0025] The selection unit 111 selects an optimal operating condition of the robot 2 by determining a condition relating to the workpiece 71 and / or a condition relating to the robot 2 from the machine learning device 15 and applying the determined condition(s) to a learning model stored in the storage unit 12.

[0026] The condition relating to workpiece 71 includes, for example, the feed amount of workpiece 71 guided by the workpiece feed device 4, the position of workpiece 71 as fed by the workpiece feed device 4, the position in the workpiece discharge device 5 to which the workpiece 71 is discharged, and / or the position of the workpiece 71 being discharged.

[0027] The selection unit 111 can select the operating condition by applying a first time period and / or a second time period to the learning model obtained from the machine learning device 15. The first time period is the time period required for the first workpiece 71a to reach the fault line 63, which specifies the lower limit before which the robot 2 will not fail to pick up the first workpiece 71a, within the workpiece picking possibility area 60. The second time period is the time period required for the second workpiece 71b to reach the fault line 63, which specifies the lower limit before which the robot 2 will not fail to pick up the second workpiece 71b, within the workpiece picking possibility area 60.

[0028] This allows selection unit 111 to select an operating condition that specifies a faster operating speed for picking up the first workpiece 71a with a decreasing first time period. Selection unit 111 can select an operating condition that specifies a faster operating speed for placing the first workpiece 71a with a decreasing second time period. Selection unit 111 can select an operating condition that specifies a faster operating speed for a picking or placing movement for the first workpiece 71a, with a decreasing difference between the first and second time periods.

[0029] The condition relating to robot 2 can include the current position of robot 2 and / or the position of robot 2.

[0030] The operating control unit 112 controls the operation of the robot 2 according to the operating condition selected by the selection unit 111.

[0031] The monitoring unit 113 monitors information about robot 2, such as the actual position of robot 2 and the position of robot 2. In addition, the monitoring unit 113 constantly monitors the position and orientation of workpiece 71 via the detection device 3. In particular, the monitoring unit 113 can constantly monitor the positions and orientations of the first workpiece 71a, which is to be picked up first from among the several workpieces fed by the workpiece feeder 4, and the second workpiece 71b, which is to be picked up next after the first workpiece, via the detection device 3.

[0032] The change amount calculation unit 114 calculates a change amount of the axes of the robot 2, based on the position and orientation of the workpiece 71 monitored by the monitoring unit 113.

[0033] In particular, the change amount calculation unit 114 can perform the change amount of the axes of robot 2, which performs the pick-up movement for the first workpiece 71a, based on the position and orientation of the first workpiece 71a and the actual position and orientation of robot 2. In this case, as a result of applying the change amount to the learning model, the selection unit 111 can select an operating condition that specifies a higher operating speed for the pick-up movement for the first workpiece 71a with increasing change amount.

[0034] Alternatively, the change amount calculation unit 114 can calculate a change amount of the axes of robot 2, which performs the placement movement for the first workpiece 71a, based on the position and orientation of the first workpiece 71a and the position and orientation of robot 2, which performs the movement to place the first workpiece 71a on the workpiece removal device 5. In this case, as a result of applying the change amount to the learning model, the selection unit 111 can select an operating condition that specifies a higher operating speed for the placement movement for the first workpiece 71a with increasing change amount.

[0035] Alternatively, the change amount calculation unit 114 can calculate a change amount of the axes of the robot 2, which performs the pick-up movement for the second workpiece 71b, based on the position and orientation of the movement for placing the first workpiece 71a on the workpiece removal device 5 and the position and orientation of the second workpiece 71b. In this case, as a result of applying the change amount to the learning model, the selection unit 111 can select an operating condition that specifies a higher operating speed for the pick-up or placement movement for the first workpiece 71a with increasing change amount.

[0036] The electrical power amount calculation unit 115 calculates an amount of electrical current consumed per unit time period when the workpiece transfer system 1 continuously transfers several workpieces 71.

[0037] The feature amount output unit 116 outputs as a feature amount the information relating to the robot, such as the actual position and the position of the robot 2, the information indicating the position and position of the workpiece 71, which is obtained as a result of monitoring carried out by the monitoring unit 113, and the information indicating the change amount of the axes of the robot 2, which is obtained as a result of a calculation carried out by the change amount calculation unit 114, for example to the machine learning device 15.

[0038] The storage unit 12 stores the learning model obtained from the machine learning device 15.

[0039] Fig. Figure 3 is a block diagram showing details of the control device 10 and the machine learning device 15 in the workpiece transfer system 1. For the sake of clarity, it shows Fig. 3 only the feature amount output unit 116 from the components of the control device 10.

[0040] The machine learning device 15 performs supervised learning to construct a learning model that is to be used by the control device 10 to specify an operating condition of the robot 2. For this purpose, as in Fig. As shown in Figure 3, the machine learning device 15 includes an input data detection unit 16, a marker detection unit 17, a learning unit 18 and a learning model storage unit 19.

[0041] The input data determination unit 16 determines a characteristic amount from the characteristic amount output unit 116 of the control device 10.

[0042] The marking detection unit 17 determines a marking that indicates whether the processing power of robot 2 is excessive or insufficient.

[0043] The state in which "the processing power of robot 2 is insufficient" here means a state in which the input quantity of workpieces 71 is excessive relative to the processing power of robot 2. In this state, robot 2 might, for example, miss picking up a workpiece 71.

[0044] On the other hand, the state in which "the processing power of robot 2 is excessive" here means a state in which the processing power of robot 2 is excessive relative to the amount of workpieces 71 being fed in. In this state, a standby time of robot 2 may occur, for example, until a workpiece 71 is fed into the workpiece handling area 60.

[0045] Learning unit 18 performs supervised learning based on training data constructed from a set of feature values ​​and a marker, in order to construct a learning model that is to be used by the control device 10 to specify an operating condition of the robot 2. Learning unit 18 sends the learning model constructed in this way to the control device 10. The learning model storage unit 19 stores the learning model constructed by learning unit 18.

[0046] For example, learning unit 18 can be built using a support vector machine (hereinafter also referred to as SVM). Since SVM is a well-known technology, a detailed description of SVM is not given here. SVM is a classification technique that employs supervised learning (learning based on given training data, which consists of both correct and incorrect data) and is known as a learning model with excellent classification performance. SVM is known for demonstrating high classification performance (high generalization performance) on, for example, untrained data.

[0047] Learning Unit 18 uses a digitized marker, as described above, which indicates whether the processing power of Robot 2 is excessive or insufficient. Additionally, Learning Unit 18 calculates a hyperplane in which a space containing the feature values ​​described above is separated by a maximum margin, based on whether the processing power of Robot 2 is excessive or insufficient. Learning Unit 18 can use the coefficient of the hyperplane as a parameter of the learning model, which is used by Control Device 10 to specify an operating condition for Robot 2. Learning Unit 18 outputs the learning model thus constructed to Control Device 10.

[0048] As described above, the learning model storage unit 19 stores the learning model constructed by the learning unit 18.

[0049] Referring to Fig. Sections 4 to 6B describe the following as an example of a learning model construction procedure, which is carried out by learning unit 18.

[0050] Fig. Figure 4 is a view showing coordinates to be used by Learning Unit 18 to construct the learning model. A target pickup position of the first workpiece 71a is defined by (X P , Y P , R P ). A “workpiece picking-up movement distance” is a distance by which the hand 23 moves to pick up the first workpiece 71a and is represented by D Prepresented. A “workpiece placement movement distance” is a distance by which hand 23 moves until hand 23 places the first workpiece 71a on the pallet 51 after picking up the first workpiece 71a and is represented by D D represented. A “placement position” is a position at which the robot 2 places the first workpiece 71a with hand 23 and is marked by (X D , Y D , R D). The actual position of the workpiece (i.e., a new first workpiece) 71b, which is in the most downstream position at the moment when robot 2 places the first workpiece 71a, is represented by (X1, Y1, R1). The actual position of a workpiece (i.e., a new second workpiece) 71c, which is the most downstream position at the moment when robot 2 places the first workpiece 71a, is represented by (X2, Y2, R2). The actual position of robot 2 at the moment when robot 2 places the first workpiece 71a with hand 23 is represented by (X S , Y S , R S ) represents. A position of the upper limit of the pickup area for workpiece 71 is defined by L U represented and a position of the lower limit of the pickup possibility area of ​​the workpiece 71 is defined by L L The position of a target acquisition line for workpiece 71 is represented by L. PThe speed at which the workpiece 71 is fed through the workpiece feed device 4 is represented by V. More specifically, in a case where the workpiece feed device 4 is a belt conveyor, the speed of movement of the conveyor is represented by V.

[0051] Fig. Figures 5A to 5C show a flowchart of an example of a process in which the workpiece transfer system 1 constructs a learning model.

[0052] In step S1, the control unit 11 draws the actual position (X S , Y S , R S ) of the robot 2 at the moment when the robot 2 places a workpiece 71 on the workpiece removal device 5.

[0053] In step S2, the control unit 11 records the actual position (X1, Y1, R1) of a new first workpiece 71b at the moment when the robot 2 places the workpiece 71 on the workpiece removal device 5.

[0054] In step S3, the control unit 11 records the actual position (X2, Y2, R2) of a new second workpiece 71c at the moment when the robot 2 places the workpiece 71 on the workpiece removal device 5.

[0055] In step S4, the control unit 11 calculates a time period (T1) that the first workpiece 71b needs to reach the target workpiece pickup line (L P ) to achieve, according to the following formula (1) using the speed (V) of the workpiece feed device 4 and the actual position of the first workpiece 71b. T1 = (X1 − Lp) / V

[0056] In step S5, the control unit 11 calculates a time period (T2) that is needed for the second workpiece 71 to reach the upper limit (L). U) of the workpiece holding possibility range 60, according to the following formula (2) using the speed (V) of the workpiece feed device 4 and the actual position of the second workpiece 71c. T2 = (X1 − LP) / V

[0057] In step S6, the control unit 11 sets the target pickup position of the first workpiece 71b using the following formula (3). XP = LP YP = Y1 RP = R1

[0058] In step S7, the control unit 11 calculates the movement distance (D). P ) of robot 2, which is in the picking motion for the first workpiece 71b, according to the following formula (4) using the actual position (X S , Y S , R S ) of robot 2 and the target acquisition position (X P , Y P , R P ) of the first workpiece 71b. DP=√((XP−XS)2+(YP−YS)2)

[0059] In step S8, the control unit 11 selects a condition from an operating condition table that serves as the learning model, according to which the robot determines the movement distance D. P during the time period (T1) required for the pickup movement. Additionally, robot 2 actually performs a pickup movement.

[0060] Here, the "operating conditions table" refers to a table that describes the operating conditions of robot 2, and it is stored in memory unit 12. Fig. Figure 6A shows an example of the operating conditions table.

[0061] In the operating conditions table, which serves as an example in Fig. As shown in Figure 6A, each row shows a movement speed per time period required for the recording movement, while each column shows a movement speed per movement distance (D). P) shows. The selection unit 111 selects from the operating conditions table a movement speed at the intersection of the time period (T1) required for the pickup movement and the movement distance (D). P ) out of.

[0062] In step S9, the control unit 11 measures a time period (T P ), which it needs for the actual picking movement, and calculates a target time period (TD) for a placement movement according to the following formula (5). TD=(T2−TP)

[0063] In step S10, the control unit 11 determines the position (X P ', R P '), at which the first workpiece 71b is actually picked up.

[0064] In step S11, the control unit 11 calculates a distance (D P ') between the pickup position of the first workpiece 71b and the target pickup position DP'=XP'−XP

[0065] In step S12, if DP ' within a range of a permissible lower limit of values ​​(A MIN ) up to and including a permissible upper limit of values ​​(A MAX ) (A MIN ≤ D P ' ≤ A MAX ) (Yes in S12), the process moves on to step S13. If, on the other hand, D P ' smaller than the permissible lower limit of values ​​(A) MIN ) is or the permissible upper limit of values ​​(A MAX If the limit is exceeded, the process advances to step S14.

[0066] In step S13, the control unit 11 determines that the value of the operating speed, selected from the operating condition table, is appropriate and updates the learning model based on the set of feature amount entered by the input data determination unit 16 and the mark indicating that the processing power of the robot 2 is appropriate.

[0067] In step S14, learning unit 18 updates the learning model based on the set of feature values ​​entered by input data determination unit 16 and the marker indicating whether the processing power of robot 2 is excessive or insufficient. As a result of updating the learning model, learning unit 18 can, for example, subdivide the operating conditions table.

[0068] Fig. 6B shows an example of a table that is a result of subdividing the operating conditions table, which is in Fig. 6A is shown, and is obtained. In the example in Fig. In table 6B, the column “180 < D” is used. P ≤ 220“ in the as an example in Fig. Table 6A is divided into the column “180 < D P ≤ 200” and the column “200 < D” P ≤ 220".

[0069] In step S15, the control unit 11 determines the placement position (X D , Y D , R D ), in which the first workpiece 71b is placed.

[0070] In step 16, the control unit 11 calculates the movement distance (D). D ) of the robot, which is in the placement movement for the first workpiece 71b, according to the following formula (7) using the pick-up position of the first workpiece 71b and the placement position of the first workpiece 71b. DD=√((XD XP`)2+(YD−YP`)2)

[0071] In step S17, the control unit 11 selects a condition from the operating condition table, according to which the robot determines its movement distance (D). DP ) in the time period (T D ), which is needed for the placement movement. In addition, robot 2 actually carries out the placement movement.

[0072] In step S18, at the end of the placement movement for the first workpiece 71b, the second workpiece 71c becomes a new first workpiece. The control unit 11 determines the actual position (X1, Y1, R1) of the first workpiece (Ex-second workpiece) 71c at the time the placement movement is completed.

[0073] In step S19, the control unit 11 calculates a distance (D1) between the first workpiece 71c and the upper limit (L U ) of the workpiece holding possibility range 60 according to the following formula (8). D1=X1−LU

[0074] In step S20, if D1 is within the range of the permissible lower limit of values ​​(A MIN ) up to the permissible upper limit of values ​​(A MAX ) including is (A MIN ≤ D1 ≤ A MAX ) (Yes in S20), the process advances to step S21. On the other hand, if D1 is smaller than the permissible lower limit (A MIN ) is or the permissible upper limit of values ​​(AMAX If this value is exceeded, the process advances to step S22.

[0075] In step S21, the control unit 11 determines that the operating speed value selected from the operating condition table is appropriate and updates the learning model based on the set of feature values ​​input by the input data determination unit 16 and the marker indicating that the processing power of robot 2 is appropriate. The process then returns to step S1.

[0076] In step S22, the control unit 11 updates the learning model based on the set of feature values ​​input by the input data determination unit 16 and the marker indicating whether the processing power of robot 2 is excessive or insufficient. As a result of updating the learning model, the learning unit 18 can, for example, subdivide the operating condition table. The process then returns to step S1.

[0077] In the configuration described above, in which the control unit 11 is configured to select an optimal operating condition of the robot 2 from the operating condition table, which serves as the learning model, based on the condition relating to the workpiece 71 and / or the condition relating to the robot 2, if a multitude of operating conditions are determined as a result of the selection, the control unit 11 can select an operating condition that results in a smaller amount of electrical current consumed per unit time period than any other of the multitude of operating conditions.

[0078] With the control device 10 having the configuration described above, the workpiece transfer system 1 according to the present embodiment can not only cause the robot 2 to operate according to the optimal operating condition, but also change the configuration of the operating condition recorded in the storage unit 12 based on the operating state of the robot 2.

[0079] A workpiece transfer system 1 according to the present embodiment comprises: a workpiece feeder 4 configured to feed at least one workpiece 71; a robot 2 configured to transfer the fed workpiece 71; a workpiece discharge device 5 on which the at least one workpiece 71 transferred from the workpiece feeder 4 by the robot 2 is to be placed; a machine learning device 15 configured to construct a learning model used to specify an operating condition of the robot 2; and a control device 10 configured to apply a condition relating to the workpiece 71 and / or a condition relating to the robot 2 to the learning model and to control the robot 2 according to the operating condition.wherein the machine learning device 15 includes an input data detection unit 16 configured to determine as input data a correspondence relationship between the condition relating to the at least one workpiece 71 and / or the condition relating to the robot 2 and the operating condition, a marker detection unit 17 configured to determine as marker information indicating whether the processing power of the robot 2 is excessive or insufficient, and a learning unit 18 configured to perform supervised learning based on training data constructed from a set of input data and the marker to construct the learning model to be used to specify the operating condition of the robot 2, and the control device 10 includes a storage unit 12,in which the learning model obtained from the machine learning device 15 is stored, a selection unit 111, which is configured to select an optimal operating condition for the robot 2 based on the learning model, and an operating control unit 112, which is configured to control the operation of the robot 2 according to the operating condition selected by the selection unit 111.

[0080] With the above configuration, the robot's operating condition is controlled according to the specified condition. This can have the effect of avoiding acceleration / deceleration of robot 2, which might otherwise be necessary due to a wasteful standby state. Additionally, with the above configuration, robot 2 operates at a necessary and sufficient operating speed. This can result in reduced electrical power consumption and extend the lifespan of the speed reducer.

[0081] Furthermore, with the above configuration it is possible to flexibly change or update the operating pattern and / or the operating condition based on the workpiece feeding status and / or the operating status of the robot.

[0082] The following describes a second embodiment of the present invention with reference to Fig. 7. For ease of description, components identical to those of the first embodiments are described with identical reference numerals to those of the first embodiment, and the functions of such components are not described here.

[0083] Fig. Figure 7 is a view showing an overall configuration of a workpiece transfer system 1A according to the second embodiment of the present invention.

[0084] According to workpiece transfer system 1A, workpieces 71 are randomly arranged on a workpiece feeder 4. In this respect, workpiece transfer system 1A differs from workpiece transfer system 1, in which the workpieces 71 are arranged in a row. To achieve this, workpiece transfer system 1A includes a robot 2A, which differs from the robot 2 operating in workpiece transfer system 1. Robot 2A includes an arm 22A instead of arm 22 and a hand 23A instead of hand 23. In these respects, robot 2A differs from robot 2.

[0085] The second embodiment is based on the assumption that the speed of a conveyor provided at the workpiece feed device 4 is lower than that of the first embodiment and that the time period required for a workpiece 71 to pass the workpiece receiving area 60 is longer than that of the first embodiment.

[0086] Fig. Figure 8 is a view showing a configuration of the workpiece holding area 60 according to the second embodiment. The workpiece holding area 60 includes a “central flow area” corresponding to a workpiece removal device 5, a “current up area” located upstream of the central flow area, and a “current down area” located downstream of the central flow area.

[0087] The machine learning device 15A according to the second embodiment includes a marking detection unit 17A instead of the marking detection unit 17. In this respect, the machine learning device 15A differs from the machine learning device 15 according to the first embodiment. If the robot 2 continuously picks up the workpieces 71 from the "current-up area," the marking detection unit 17A determines that the processing performance is excessive. If the robot 2 continuously picks up the workpieces 71 from the "medium-current area," the marking detection unit 17A determines that the processing performance is adequate. If the robot 2 continuously picks up the workpieces 71 from the "current-down area," the marking detection unit 17A determines that the processing performance is insufficient. The learning unit 18 constructs a learning model using such a marking.

[0088] The workpiece transfer system 1A according to the present embodiment is configured as below. If the robot 2 continuously picks up the workpiece 71 from the "current-up area" in the workpiece pick-up area 60, the marking detection unit 17A determines that the processing performance is excessive. If the robot 2 continuously picks up the workpiece 71 from the "mid-current area" in the workpiece pick-up area 60, the marking detection unit 17A determines that the processing performance is adequate. If the robot 2 continuously picks up the workpiece 71 from the "current-down area" in the workpiece pick-up area 60, the marking detection unit 17A determines that the processing performance is insufficient.

[0089] With the above configuration, especially in a case where the speed of the conveyor of the workpiece feeding device 4 is low and the time period that the workpiece 71 needs to pass through the workpiece holding possibility area 60 is long, it is possible to flexibly change or update the operating pattern and / or the operating condition based on the feeding state of the workpiece 71 and / or the operating state of the robot 2. EXPLANATION OF REFERENCE MARKS 1, 1A Workpiece Transfer System 2, 2A Robot 4. Workpiece feed device 5 Workpiece removal device 10, 10A control device 11 Control unit 12 storage units 15, 15A Machine learning device 16 Input data determination unit 17, 17A Marking Determination Unit 18 Learning Unit 19 Learning Model Storage Unit 60 workpiece holding option range 71 workpiece 111 Selection Unit 112 Operating control unit 113 Monitoring Unit 114 Change Amount Calculation Unit 115 Electrical power amount calculation unit 116 Feature amount output unit

Claims

[1] Workpiece transfer system (1), comprising: a workpiece feeding device (4) configured to feed at least one workpiece (71); a robot (2) configured to transfer the at least one workpiece (71) being fed; a workpiece removal device (5) on which the at least one workpiece (71) transferred by the robot (2) from the workpiece feed device (4) is to be placed; a machine learning device (15) configured to construct a learning model used to indicate an operating condition of the robot (2); and a control device (10) configured to select the operating condition of the robot (2) by applying a condition relating to the at least one workpiece (71) and / or a relationship relating to the robot (2) to the learning model, and to control the robot (2) according to the operating condition, wherein the machine learning device (15) includes an input data determination unit (16) configured to determine, as input data, a correspondence relationship between the condition relating to the at least one workpiece (71) and / or the condition relating to the robot (2) and the operating condition, a mark determination unit (17) configured to determine, as a mark, information indicating whether or not a processing power of the robot (2) is excessive, or whether or not the processing power of the robot (2) is insufficient, and a learning unit (18) configured to perform supervised learning based on teaching data constructed by a set of the input data and the marker to construct the learning model to be used to indicate the operating condition of the robot (2), and the control device (10) includes a storage unit (12) in which the learning model obtained from the machine learning device (15) is stored, a selection unit (111) configured to select an optimal operating condition as the operating condition of the robot based on the learning model, and an operation control unit (112) configured to control the operation of the robot according to the operation condition selected by the selection unit (111). [2] The workpiece transfer system (1) according to claim 1, wherein the condition relating to the at least one workpiece (71) is a condition relating to a feed amount of the at least one workpiece (71), a position of the at least one workpiece (71) being fed, a position in the workpiece discharge device (5) to which position the at least one workpiece (71) is discharged, and / or a position of the at least one workpiece (71) being discharged. [3] Workpiece transfer system (1) according to claim 1 or 2, wherein the condition relating to the robot (2) is a condition relating to an actual position of the robot (2) and / or a posture of the robot (2). [4] Workpiece transfer system (1) according to one of claims 1 to 3, wherein the control device (10) further includes a monitoring unit (113) configured to constantly monitor a position and a posture of a first workpiece to be picked up from a plurality of workpieces (71) fed by the workpiece feeding device (4), and a position and a posture of a second workpiece to be picked up next to the first workpiece from the plurality of workpieces (71), wherein the selection unit (111) is further configured to select the operating condition by applying a first time period and / or a second period to the learning model, wherein the first time period is a time period required for the first workpiece to reach a lower limit line in a workpiece pick-up possibility range set in the workpiece supply device (4), before which the robot (2) does not fail to pick up the first workpiece, wherein the second time period is a time period required for the second workpiece to reach the lower limit line in the workpiece pick-up possibility range, before which the robot (2) does not fail to pick up the second workpiece. [5] The workpiece transfer system (1) according to claim 4, wherein the selection unit (111) is further configured to select, as the operating condition, an operating condition indicating a faster speed for picking up the first workpiece as the first time period decreases, the selection unit is further configured to select, as the operating condition, an operating condition indicating a faster speed for placing the first workpiece as the second time period decreases, and the selection unit is further configured to select, as the operating condition, an operating condition indicating a faster speed for a picking movement or a placing movement for the first workpiece as a difference between the first time period and the second time period decreases. [6] Workpiece transfer system (1) according to one of claims 1 to 3, wherein the control device (10) further includes a monitoring unit (113) configured to constantly monitor a position and a posture of the at least one workpiece, and a change amount calculation unit (114) configured to calculate a change amount of axes of the robot based on the position and posture of the at least one workpiece monitored by the monitoring unit, and the selection unit (111) is further configured to select the operating condition by applying the change amount to the learning model. [7] The workpiece transfer system (1) according to claim 6, wherein the monitoring unit (113) is further configured to constantly monitor a position and a posture of a first workpiece to be picked up first among the plurality of workpieces fed by the workpiece feeding device, the change amount calculation unit (114) is further configured to calculate a change amount of the axes of the robot (2) performing the pick-up movement for the first workpiece based on a position and a posture of the first workpiece and an actual position and posture of the robot (2), and the selection unit (111) is further configured to select the operating condition by applying the change amount to the learning model. [8] The workpiece transfer system (1) according to claim 7, wherein the selecting unit (111) is further configured to select, as the operating condition, an operating condition indicating a higher operating speed for the pick-up movement for the first workpiece with increasing change amount. [9] Workpiece transfer system (1) according to one of claims 6 to 8, wherein the monitoring unit (113) is further configured to constantly monitor the position and posture of the first workpiece to be picked up first from the plurality of workpieces (71) fed by the workpiece feeding device (4), the change amount calculation unit (114) is further configured to calculate a change amount of the axes of the robot (2) performing the placing movement for the first workpiece based on the position and posture of the first workpiece and a position and posture of the robot (2) performing a movement for placing the first workpiece on the workpiece removal device, and the selection unit (111) is further configured to select the operating condition by applying the change amount to the learning model. [10] The workpiece transfer system (1) according to claim 8, wherein the selecting unit (111) is further configured to select, as the operating condition, an operating condition indicating a higher operating speed for the placing movement for the first workpiece with increasing change amount. [11] The workpiece transfer system (1) according to any one of claims 7 to 9, wherein the monitoring unit (113) is further configured to constantly monitor a position and a posture of a second workpiece to be picked up next after the first workpiece from the plurality of workpieces (71) fed by the workpiece feeding device (4), the change amount calculation unit (114) is further configured to calculate a change amount of the axes of the robot that performs a pick-up movement for the second workpiece, based on a position and a posture of the robot (2) that performs a movement for placing the first workpiece on the workpiece removal device, and the position and posture of the second workpiece, and the selection unit (111) is further configured to select the operating condition by applying the change amount to the learning model. [12] The workpiece transfer system (1) according to claim 11, wherein the selecting unit (111) is further configured to select, as the operating condition, an operating condition indicating a higher operating speed for the picking movement or the placing movement for the first workpiece with increasing change amount. [13] Workpiece transfer system (1A) according to one of claims 1 to 3, wherein the at least one workpiece comprises a plurality of workpieces, in a case where the robot (2) continuously picks up the plurality of workpieces (71) from a power-on area in a workpiece pick-up possibility area (60) set in the workpiece supply device (4), wherein the mark detection unit (17A) determines that the processing power is excessive, in a case where the robot (2) continuously picks up the plurality of workpieces (71) from a mid-stream area in the workpiece pick-up capability area (60), the mark detection unit (17A) determines that the processing performance is appropriate, and in a case where the robot continuously picks up the plurality of workpieces (71) from a power-down area in the workpiece pick-up possibility area (60), the mark detection unit (17A) determines that the processing performance is insufficient. [14] The workpiece transfer system (1, 1A) according to any one of claims 1 to 3 and 13, wherein in a case where a standby time of the robot (2) occurs until the at least one workpiece (71) is supplied to the workpiece receiving possibility area (60) set in the workpiece supply device (4), the mark detection unit (17, 17A) determines that the processing power is excessive. [15] The workpiece transfer system (1, 1A) according to any one of claims 1 to 3, 13 and 14, wherein in a case where the robot (2) fails to pick up the at least one workpiece (71), the mark detecting unit (17, 17A) determines that the processing performance is insufficient. [16] Workpiece transfer system (1) according to one of claims 1 to 3, wherein the control device (10) further includes an electric power amount calculation unit (115) configured to calculate an amount of electric power consumed per unit time period in a case where workpieces are continuously transferred, and the selection unit (111) is further configured to select, in a case where a plurality of operating conditions are determined as a result of selecting the optimal operating condition, an operating condition that results in a smaller amount of electric power consumed per unit time period than any other of the plurality of operating conditions.

Images