WO2019239565A1

WO2019239565A1 - Component serving apparatus for kitting tray

Info

Publication number: WO2019239565A1
Application number: PCT/JP2018/022811
Authority: WO
Inventors: 健次石塚; 亮介中村; 智仁内海
Original assignee: ヤマハ発動機株式会社
Priority date: 2018-06-14
Filing date: 2018-06-14
Publication date: 2019-12-19
Also published as: CN112218746A; CN112218746B; JPWO2019239565A1; JP7011063B2

Abstract

A component serving apparatus (1) for serving a plurality of types of components having different sizes to a kitting tray (40) provided with a plurality of storage sections, is provided with a robot hand (10) and a control unit (30) that controls operation of the robot hand (10). The robot hand (10) has a head part (15) capable of picking and releasing components, performs picking of a target component at a preservation location of the components with use of the head part (15) and conveys the target component, and performs serving operation for releasing the target component from the head part (15) to the kitting tray (40). The control unit (30) is provided with: a rule setting unit that, in accordance with modes of the plurality of storage sections and the plurality of types of components, sets serving rules of serving the components to the kitting tray (40); and a drive control unit (31) that, on the basis of the serving rules, causes the robot hand (10) to execute the serving operation.

Description

Parts distribution device for kitting tray

The present invention relates to a component arrangement apparatus that arranges a plurality of types of parts having different sizes on a kitting tray provided with a plurality of storage units.

The kitting tray is a tray that accommodates a set of component groups including fastening parts such as screws, bolts, and washers, and sealing parts such as O-rings and gaskets used when assembling one machine product. Japanese Patent Application Laid-Open No. 2004-151561 discloses picking of a part from such a kitting tray by a robot hand, and application of machine learning for accurately performing the picking.

The kitting tray has a plurality of storage compartments separated by a frame. In these storage compartments, parts are arranged from storage trays for storing various parts. A large part that does not enter the housing compartment may be placed on the upper edge of the frame or leaned up. It is desirable to automate the arrangement of parts on the kitting tray using a robot hand. Japanese Patent Application Laid-Open No. H10-228867 only discloses the technology for picking parts, and does not mention the arrangement of parts to the kitting tray.

JP 2017-30135 A

An object of the present invention is to provide a component arrangement device capable of accurately arranging various components on a kitting tray using a robot hand.

A component arrangement device for a kitting tray according to one aspect of the present invention is a component arrangement device that arranges a plurality of types of components of different sizes on a kitting tray having a plurality of storage units, and picking and releasing the components. An arrangement having a possible head part, picking and transporting a target part from the plurality of types of parts at the storage position of the part by the head part, and releasing the target part from the head part to the kitting tray A robot hand that performs an operation; and a control unit that controls the operation of the robot hand, wherein the control unit is configured to control the kitting tray of the component according to the types of the plurality of components and the plurality of storage units. A rule setting unit for setting a distribution rule to the robot, and causing the robot hand to execute the distribution operation based on the distribution rule Comprising a drive control unit.

FIG. 1 is a block diagram illustrating a configuration of a component layout device for a kitting tray according to an embodiment of the present invention. FIG. 2A is a top plan view showing an example of a kitting tray on which components are arranged, and FIG. 2B is a cross-sectional view taken along the line IIB-IIB in FIG. FIG. 3 is a flowchart showing the operation of the component layout apparatus. 4A to 4D are diagrams for explaining an example of a layout rule. FIGS. 5A and 5B are diagrams for explaining an example of a layout rule. FIGS. 6A to 6D are diagrams for explaining an example of a layout rule. FIGS. 7A to 7D are diagrams for explaining an example of the layout rule. FIGS. 8A to 8C are diagrams for explaining an example of a layout rule. FIG. 9 is a flowchart illustrating an example of the learning operation of the layout rule. FIG. 10 is a flowchart illustrating an example of the learning operation of the layout rule. FIGS. 11A and 11B are diagrams for explaining an example of a layout rule.

[Overall configuration of component arrangement equipment]
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. First, with reference to the block diagram of FIG. 1, the structure of the component arrangement | positioning apparatus 1 to the kitting tray which concerns on embodiment of this invention is demonstrated. The component arrangement apparatus 1 is an apparatus that arranges a plurality of types of parts of different sizes on the kitting tray 40, and includes a robot hand 10 that performs an arrangement operation of the parts, and a three-dimensional measurement apparatus 20 that captures a three-dimensional image of the parts ( An imaging device) and a control unit 30 that comprehensively controls the operations of the robot hand 10 and the three-dimensional measuring device 20.

The robot hand 10 is a robot apparatus that picks a target part from a part storage tray Ta, Tb,... That stores a plurality of types of parts Wa, Wb,. The component storage trays Ta, Tb,... Are trays that individually store the components Wa, Wb,..., And various components Wa, Wb,. . The kitting tray 40 is a tray that can individually store a plurality of types of components Wa, Wb, Wc, Wd,... In a plurality of storage sections partitioned in advance. The robot hand 10 picks and carries the target part, and performs a catering operation to release the target part picked in the housing portion.

The robot hand 10 is a multi-axis multi-joint robot including a base part 11, a body part 12, a first arm 13, a second arm 14, a head part 15, and a hand part 16. The base unit 11 is a housing that is fixedly installed on a floor surface, a pedestal, or the like. The body portion 12 is disposed on the upper surface of the base portion 11 so as to be rotatable in both forward and reverse directions around the first shaft 1A extending vertically. The first arm 13 is an arm member having a predetermined length, and a base end portion of the first arm 13 is attached to the trunk portion 12 so as to be rotatable around an axis of the second shaft 1B extending in the horizontal direction. The second arm 14 is an arm member that is connected to the first arm 13, and a base end portion of the second arm 14 is pivotable around the third axis 1 </ b> C extending in the horizontal direction. It is attached.

The head portion 15 is attached to the distal end side of the second arm 14 so as to be rotatable around the axis of the fourth axis 1D extending in the horizontal direction. The hand unit 16 is attached to the head unit 15 via a fifth shaft 1E perpendicular to the fourth shaft 1D. The hand unit 16 can pick the parts Wa, Wb... Individually from the part storage trays Ta, Tb... And release them to the kitting tray 40. The hand portion 16 is composed of a pair of gripping pieces, is rotatable around the axis of the fifth shaft 1E, and an interval between the pair of gripping pieces is adjustable for the picking and releasing. Note that it is only necessary that the head unit 15 has a mechanism capable of picking and releasing components. For example, instead of the hand unit 16, an electromagnet that generates an attractive force for the component or a negative pressure generator is used as the head. The unit 15 may be provided.

The three-dimensional measuring apparatus 20 includes a first camera 21, a second camera 22, and a camera control unit 23. The first camera 21 is disposed above the component storage trays Ta, Tb,..., And captures an image including the components Wa, Wb,. The second camera 22 is disposed above the kitting tray 40 and captures an image including the parts Wa, Wb, Wc, and Wd stored in the kitting tray 40. Note that the three-dimensional measuring device 20 or the first and

second cameras

21 and 22 may be mounted on the robot hand 10.

The camera control unit 23 causes the first and

second cameras

21 and 22 to perform an imaging operation, and performs three-dimensional measurement of parts based on the obtained image. The camera control unit 24, the image processing unit 25, including. The imaging control unit 24 causes the first camera 21 to perform an operation of imaging the component storage trays Ta, Tb... When picking the components Wa, Wb. Further, the imaging control unit 24 performs an operation of imaging the kitting tray 40 on the second camera 22 when confirming the position of the storage unit of the kitting tray 40 or when acquiring information on the kitting tray 40 after the parts are arranged. Let it run.

The image processing unit 25 performs image processing on the images acquired by the first and

second cameras

21 and 22 to generate image data including the three-dimensional position information of each component. The three-dimensional position information of each part is represented by coordinate values (X, Y, Z) using an XYZ orthogonal coordinate system, for example. The position information of the components Wa, Wb... Stored in the component storage trays Ta, Tb. This position information is used when the parts Wa, Wb... Are picked by the robot hand 10. Further, the position information of the parts Wa, Wb, Wc, Wd arranged on the kitting tray 40 is acquired from the image acquired by the second camera 22. Based on this position information, the component arrangement state can be evaluated.

The control unit 30 includes a drive control unit 31, a rule setting unit 32, an information acquisition unit 33, and a learning unit 34. The drive control unit 31 causes the robot hand 10 to execute a component arranging operation based on the arrangement rule set by the rule setting unit 32. The drive control unit 31 operates a drive motor (not shown) provided in the robot hand 10 so as to sequentially execute picking of parts, holding and transporting of the parts, and release of the parts in accordance with the layout rules. To control. In the learning unit 34, when machine learning regarding the caulking operation is executed, information on how the drive control unit 31 operated the robot hand 10 in the picking and release is output to the learning unit 34.

The rule setting unit 32 determines the size of the parts Wa, Wb... According to the size and shape of the parts Wa, Wb. A rule for the arrangement to the kitting tray 40 is set. This arrangement rule is an agreement regarding the arrangement concept such as matters to be prioritized and items to be avoided when parts are arranged on the kitting tray 40. For example, the layout rule is a rule in which small-sized parts are preferentially arranged in a small accommodating portion, and when parts are arranged so as to overlap in the vertical direction, which parts are arranged first (FIGS. 4 to 5). 8 will be described later in detail). This layout rule may be a layout rule taught by the operator via the input unit 26 or the like, or a layout rule in which the taught layout rule is evaluated and revised based on the layout result, or the learning unit 34. As a result of machine learning according to the above, it may be a rule created or revised.

The information acquisition unit 33 acquires information input by the operator from the input unit 26 and three-dimensional measurement information from the camera control unit 23. From the input unit 26, for example, attribute information such as the size and shape of parts, information on the shape of the kitting tray 40, and the like are given. From the camera control unit 23, the three-dimensional position information of the parts Wa, Wb,... In the part storage trays Ta, Tb, the three-dimensional position information of the storage part of the kitting tray 40, the parts in the kitting tray 40. The three-dimensional position information of Wa, Wb, Wc, and Wd is given. Based on various kinds of information given to the information acquisition unit 33, the rule setting unit 32 sets or revises the layout rule.

The learning unit 34 is a functional unit that executes a learning process for learning the operation of the robot hand 10. When setting the layout rule by machine learning, the learning unit 34 controls the robot hand 10 by the drive control unit 31 and the three-dimensional measurement information indicating the layout result to the kitting tray 40 input from the camera control unit 23. Are acquired for each learning cycle. Then, the learning unit 34 learns the optimum behavior pattern of the robot hand 10 when each component is arranged from these pieces of information, and reflects this in the arrangement rule. The behavior pattern includes, for example, at which position the part is gripped and picked with what gripping force, at which three-dimensional position of the kitting tray 40 the part is released, and in what order the plural kinds of parts are arranged. It is the action of the robot hand 10 regarding whether or not it is served. The learning unit 34 includes a displacement amount observation unit 35, a reward setting unit 36, and a value function updating unit 37. These will be described in detail when an embodiment to which machine learning is applied is described later.

[Kitting tray]
FIG. 2A is a top plan view showing an example of the kitting tray 40 on which parts are arranged, and FIG. 2B is a cross-sectional view taken along the line IIB-IIB in FIG. The kitting tray 40 is a tray that accommodates a set of parts such as a fastening part such as a screw, a bolt, and a washer used when assembling one machine product, and a sealing part such as an O-ring and a gasket. The kitting tray 40 is a relatively shallow rectangular parallelepiped container having an open top surface, and includes an outer frame portion 41 that forms an outer shape portion, an inner frame portion 42 disposed inside the outer frame portion 41, And a bottom plate 43 that forms the bottom surface of the kitting tray 40.

The outer frame portion 41 has a rectangular shape as viewed from above in FIG. 2A, and its upper edge 41T is located at the kitting tray 40 as shown in the sectional view of FIG. The side plate at the highest position. The inner frame portion 42 is a frame plate that extends vertically and horizontally in the outer frame portion 41 in a plan view and divides the plurality of accommodating portions A1, A2, and A3. Here, one large accommodating portion A1 having the largest accommodating space, four middle accommodating portions A2 having the next largest accommodating space, and eight small accommodating portions A3 having the smallest accommodating space are the inner frame portions. The example divided by 42 is shown. The height of the upper end edge 42T of the inner frame portion 42 is lower than the upper end edge 41T of the outer frame portion 41. Thereby, the upper stage accommodating part A0 which uses the upper end edge 42T as a bottom face and uses the outer frame part 41 as a partition wall is formed.

The same parts are arranged in each accommodating part A1, A2, A3. Here, a large-sized component W1 composed of a large-diameter seal ring is provided in the large accommodating portion A1, medium-sized components W21 and W22 such as a medium-diameter C ring in the middle accommodating portion A2, and small-sized components such as bolts and nuts in the small accommodating portion A3. The example in which the components W31 and W32 are accommodated is shown. These components W1, W21, W22, W31, and W32 are accommodated in the accommodating portions A1, A2, and A3 so as to contact the bottom plate 43. Further, an example is shown in which an ultra-large component W0 such as a gasket that does not enter any of the accommodating portions A1 to A3 is accommodated in the upper accommodating portion A0. The ultra-large component W0 is in contact with the upper end edge 42T of the inner frame portion 42.

The kitting tray 40 on which the parts W0 to W32 are arranged is transported to a predetermined work place by an operator or a transport robot. Alternatively, the parts W0 to W32 may be taken out from the kitting tray 40 by another robot hand. In addition, although the machine part was illustrated as a part here, the part accommodated in the kitting tray 40 does not have a restriction | limiting in particular. For example, power / electronic parts, material chips, material rods, tools, and the like can be accommodated in the kitting tray 40 as parts.

[Basic flow of catering operation]
FIG. 3 is a flowchart showing the basic operation of the component layout apparatus 1. First, the information acquisition unit 33 of the control unit 30 acquires information regarding the kitting tray 40 and the parts W0 to W32 (step S1). This information includes information on the shape of the kitting tray 40, the mode of the accommodating portions A1 to A3 (opening size, depth, arrangement position, etc.), the arrangement position with respect to the robot hand 10, the type, shape, size, etc. For example, the arrangement position with respect to the robot hand 10. The information acquisition unit 33 acquires the above information based on an input operation via the input unit 26 or a measurement result by the three-dimensional measurement apparatus 20.

Next, the rule setting unit 32 sets a layout rule for the target parts W0 to W32 accommodated in the kitting tray 40 based on the information acquired by the information acquisition unit 33 (step S2). Later, some specific examples of the arrangement rule will be described. As described above, the layout rule is formulated from information acquired by the information acquisition unit 33 or by machine learning in which the result of actual allocation is evaluated by measurement of the three-dimensional measuring device 20. Things can be used.

Subsequently, the drive control unit 31 drives the robot hand 10 according to the set layout rule, and executes a layout operation for distributing the target parts W0 to W32 to the storage units A1 to A3 of the kitting tray 40 (step S3). That is, the drive control unit 31 rotates the robot hand 10 about the axis of the first axis 1A and appropriately operates the second axis 1B to the fifth axis 1E, thereby moving the hand unit 16 to the component storage tray Ta, The components Wa, Wb... Are picked individually toward Tb..., Are transported to the specified accommodating portions A1 to A3 of the kitting tray 40, and are released. The drive control unit 31 confirms whether or not all scheduled parts W0 to W32 have been arranged for one kitting tray 40 (step S4), and if the arrangement has not been completed (step S4). NO in S4) continues the catering operation in step S3.

When the arrangement is completed (YES in step S4), the control unit 30 determines whether or not to perform the arrangement result evaluation (step S5). When the result evaluation is executed (YES in step S5), the information acquisition unit 33 causes the second camera 22 of the three-dimensional measurement apparatus 20 to capture the image of the kitting tray 40 after the arrangement, and the arranged parts W0 to W32 are arranged. Acquire 3D measurement results. Then, the rule setting unit 32 evaluates the layout result from the state information of the parts W0 to W32 in the kitting tray 40 based on the three-dimensional measurement result (step S6).

The rule setting unit 32 determines whether or not the current layout rule needs to be corrected based on the evaluation result (step S7). When the evaluation result is a level exceeding a predetermined threshold, the rule setting unit 32 determines that the layout rule needs to be corrected (YES in step S7), and corrects the layout rule (step S8). Examples of the correction here include correction of the picking position, grip force, grip direction, release height position, and arrangement order of the parts W0 to W32 by the hand unit 16.

On the other hand, if the layout failure is equal to or less than the threshold value, the rule setting unit 32 determines that the layout rule is not required to be modified (NO in step S7), and ends the process. In addition, when it is determined in step S5 that the evaluation of the layout result is not executed (NO in step S5), the process is also finished. Steps S5 to S8 may be executed manually by the operator or may be executed according to a predetermined correction rule. Further, the substantially same process as steps S5 to S8 can be replaced with machine learning by the learning unit 34. An example of this machine learning will be described later with reference to FIGS.

[Specific examples of serving rules]
A specific example of the layout rule set by the rule setting unit 32 will be described below with reference to FIGS. FIGS. 4A to 4D are diagrams for explaining an example of the layout rule, and are diagrams illustrating the layout rule when the layout in which the parts W11 and W12 overlap in the vertical direction is performed. is there. 4A is a top view showing a desirable arrangement state of the parts W11 and W12 to the kitting tray 40, and FIG. 4B is a side view in the direction of arrow a in FIG. 4A.

The part W11 is a ring-shaped part such as packing, and the part W12 is a long bar-shaped part such as a bolt. The parts W11 are arranged on the bottom plate 43 of the housing part A11 defined by the inner frame part 42 of the kitting tray 40 in a manner of being arranged in a matrix of 2 × 3. On the other hand, the component W12 is arranged in such a manner that it is supported by the upper end edge 42T of the inner frame portion 42 that defines the accommodating portion A11.

4A and 4B, the rule setting unit 32 sets a layout rule that allows the component W11 with the layout position below to be distributed in advance. That is, as shown in FIG. 4C, the robot hand 10 first arranges each component W11 at a predetermined position in the housing portion A11. In this case, in the case of the ring-shaped component W11, it is desirable to arrange and arrange in a matrix as illustrated in FIG. 4A so that the compartment of the housing portion A11 can be fully utilized.

When the arrangement of all the parts W11 scheduled to be arranged in the accommodating part A11 is completed, subsequently, as shown in FIG. 4D, the robot hand 10 arranges the rod-like parts W12. At this time, the component W12 is released from the hand portion 16 so that one end side and the other end side of the component W12 are respectively supported by the upper end edges 42T of the pair of inner frame portions 42 facing each other. Thereby, both components are arranged in the kitting tray 40 in such a manner that the component W12 overlaps the component W11. By adopting the layout rules of FIGS. 4C and 4D, the component W11 whose layout position should be lower is disposed on the component W12 whose layout position is higher, or the layout of the component W11. Can be prevented from being obstructed by the component W12.

FIGS. 5A, 5B, and 6A to 6D are diagrams for explaining an example of the layout rule, and minimize the overlap of components housed in the same housing portion. It is a figure which shows a layout rule. When both the arrangement in which the parts are arranged in the horizontal direction in the kitting tray 40 and the arrangement in which the parts overlap in the vertical direction can be executed, the rule setting unit 32 arranges the parts in the horizontal direction rather than the arrangement in which the parts overlap in the vertical direction. Set a distribution rule to execute with priority.

5A and 5B show an example in which six parts W21 having a hexagonal cross section are housed in the housing part A12 partitioned by the inner frame part 42. FIG. As shown in FIG. 5 (A), the six parts W21 can be arranged in a line in the horizontal direction with respect to the housing part A12, or in the vertical direction as shown in FIG. 5 (B). It can also be arranged in a stacked manner in multiple stages (two stages). That is, it is possible to adopt both a planar arrangement and a three-dimensional arrangement from the size of the component W21 and the hexagonal shape of the cross section that can be easily stacked.

In such a case, the rule setting unit 32 gives priority to the planar arrangement in FIG. In other words, it is not necessary to allow the parts W21 to be stacked in the vertical direction, but to effectively utilize the horizontal space of the housing portion A12 and perform a well-balanced arrangement. Even in a case where the number of parts W21 is large and it is necessary to arrange them in the vertical direction, as many parts W21 as possible can be arranged directly above the bottom plate 43 in the accommodating portion A12, and then two steps Have the eyes arranged. Thereby, the accommodation stability in accommodating part A12 of the some components W21 can be improved. In addition, when the components W21 are stacked, the components W21 arranged at a relatively high position from the bottom surface 43 are generated, and the components W21 are likely to jump out of the housing portion A12 when the kitting tray 40 is delivered after the arrangement. It becomes. In view of these, the planar arrangement of FIG. 5 (A) is more advantageous than the three-dimensional arrangement of FIG. 5 (B).

FIGS. 6A to 6D also show the same case as FIG. 6A is a top view showing a desirable arrangement state of the bolt-shaped parts W31 to W34 to the accommodating portion A13 of the kitting tray 40, and FIG. 6B is a view in the direction of arrow b in FIG. 6A. It is a side view. Here, the parts W31 to W34 are arranged in a line in the horizontal direction so that the bolt heads are alternately opposite to each other. By arranging the components W31 to W34 in this way, the horizontal space of the accommodating portion A13 can be effectively utilized, and the height of the arrangement of the components W31 to W34 in the accommodating portion A13 (the height of the center of gravity position) is suppressed. be able to.

On the other hand, FIG. 6 (C) is a top view showing the arrangement state of the parts W31 to W34 according to the comparative example, and FIG. 6 (D) is a side view in the direction of arrow c in FIG. 6 (C). Here, out of the four parts W31 to W34, the parts W33 and W34 are arranged on the bottom plate 43, and the parts W31 and W32 are arranged so as to be placed on the upper side thereof, thereby being arranged in a lattice manner. An example is shown. As described above, when the components W31 to W34 are arranged in the accommodating portion A13, it is possible to arrange them in a three-dimensional manner in a lattice shape instead of arranging them in a plane. However, the rule setting unit 32 sets a layout rule giving priority to planar layout as shown in FIGS. 6 (A) and 6 (B). Thereby, the housing stability of the components W31 to W34 in the housing portion A13 can be improved, and the components W31 to W34 can be prevented from popping out.

FIGS. 7A to 7D are diagrams for explaining an example of the layout rule, and are diagrams showing the layout rule for setting the layout order according to the outer area. Although the layout rule is similar to the layout rule shown in FIG. 4, the rule setting unit 32 precedes a component with a small occupation area in plan view when performing layout in which the components overlap each other in the kitting tray 40. Set the distribution rules to be distributed. The occupied area here is an area determined by an outline outline in plan view, and when there is a space in the outline outline, it is an area including the space.

FIG. 7A is a plan view showing a state in which parts W41 and W42 having different occupation areas are arranged on the kitting tray 40. FIG. The component W41 is a bolt-shaped component and is arranged in the storage portion A14 partitioned by the inner frame portion 42. That is, the component W41 has a small exclusive area that can be accommodated in one accommodating portion A14. On the other hand, the component W42 is a large ring-shaped component such as a gasket and has a large exclusive area that cannot be accommodated in any of the accommodating portions defined by the inner frame portion 42. Accordingly, the component W42 is arranged so as to be supported by the upper end edge 42T of the inner frame portion 42.

In such a case, the rule setting unit 32 arranges the component W41 having a small occupied area in advance in the storage unit A14 and then arranges the component W42 having a large occupied area. That is, as shown in FIG. 7B, first, the robot hand 10 arranges the component W41 into the storage portion A14 and contacts the bottom plate 43. If there are other parts to be accommodated in the compartments for the other accommodating parts, the other parts are also arranged. After that, as shown in FIG. 7C, the part W42 having a large occupation area is arranged at a predetermined position in contact with the upper end edge 42T of the inner frame part 42.

When the component W42 having a large occupation area is arranged in advance, the component W42 closes the upper opening of the storage portion A14. As a result, there arises a problem that it is impossible to arrange the component W41 having a small occupied area, and a problem that the component W41 is arranged on the component W42 as shown in FIG. 7D. By setting a layout rule as shown in FIGS. 7B and 7C, these problems can be prevented.

FIGS. 8A to 8C are diagrams for explaining an example of the layout rule, and are diagrams illustrating the layout rule stored in the kitting tray 40 so that the center of gravity of the component is as low as possible. The rule setting unit 32 includes a first arrangement state in which the height position of the center of gravity of the part is the first position, and the height position of the center of gravity of the part is the first position. The layout rule is set so that the component takes the first layout state when the second layout state, which is a higher second position, can be taken.

FIG. 8A is a plan view in a top view showing a state in which the parts W51 and W52 are arranged in the housing part A15 partitioned by the inner frame part 42. FIG. The parts W51 and W52 are the same bolt parts having the bolt head B1 and the bolt main body B2, and have a length that does not allow the whole part to enter the compartment of the housing portion A15. The component W51 is arranged in such a manner that the bolt head B1 side is accommodated in the accommodating portion A15 (contacting the bottom plate 43) and the bolt main body B2 side rides on the upper end edge 42T of the inner frame portion 42. Conversely, the component W52 is arranged in such a manner that the bolt body B2 side is accommodated in the accommodating portion A15 and the bolt head B1 side protrudes upward from the upper end edge 42T of the inner frame portion 42. Reference numerals G1 and G2 in the figure indicate the positions of the center of gravity of the parts W51 and W52. Since the bolt head B1 is a heavy part in the parts W51 and W52, the centers of gravity G1 and G2 are close to the bolt head B1 of the bolt body B2.

FIG. 8B is a side view showing a housing state of the component W51 in the housing portion A15. The component W51 is arranged in the accommodating portion A15 in a state where the bolt head B1 side is inclined downward and the vicinity of the tip end side of the bolt main body B2 is lifted by the upper end edge 42T (first arrangement state). The center of gravity G1 of the component W51 is located at a height h1 (first position) determined by the size of the bolt head B1 with respect to the bottom plate 43.

On the other hand, FIG. 8C is a side view showing a housing state of the component W52 in the housing portion A15. The component W52 is arranged in the housing portion A15 in a state where the bolt body B2 is inclined so that the tip of the bolt body B2 is downward and the vicinity of the root portion of the bolt body B2 near the bolt head B1 is lifted by the upper end edge 42T (second). Serving state). In this case, since the bolt head B1 side is upward, the center of gravity G2 of the component W52 is positioned at a height h2 (second position) that is higher than the center of gravity G1 of the component W51 by Δh.

In such a case, the rule setting unit 32 sets the layout rule so that the layout state adopted for the component W51 in FIG. That is, the arrangement rule is set so that the arrangement in which the height position of the center of gravity G1 is lowest can be performed in the state of being accommodated in the accommodation portion A15. As shown in FIG. 8C, in the storage mode such as the component W52 in which the center of gravity G2 exists at a relatively high position, the component W52 easily falls from the storage portion A15. On the other hand, as shown in FIG. 8 (B), by adopting a storage mode such as the component W51 in which the center of gravity G1 is lower, the component W51 can be made difficult to fall from the storage portion A15.

[Specific examples of machine learning]
Next, an example will be described in which the rule setting unit 32 sets a layout rule by machine learning by the learning unit 34 (evaluation unit). Here, an example in which the three-dimensional measuring device 20 (imaging device) acquires a three-dimensional image of the kitting tray 40 after the catering operation by the robot hand 10 is executed, and the learning unit 34 evaluates based on the three-dimensional image. Indicates.

<Configuration of learning unit>
When machine learning is executed, the learning unit 34 determines each component from the control information of the robot hand 10 when a certain arrangement operation is executed and the position information of the component in the kitting tray 40 where the arrangement operation is executed. Learns the optimal behavior pattern of the robot hand 10 when serving. The learning result acquired by the learning unit 34 is reflected in the layout rule set by the rule setting unit 32. As described above, the learning unit 34 includes a displacement amount observation unit 35, a reward setting unit 36, and a value function updating unit 37 (FIG. 1).

The displacement amount observing unit 35 includes three-dimensional image data (hereinafter referred to as basic image data) of a comparison-source kitting tray 40 on which parts are arranged, and a three-dimensional image of a comparison-target kitting tray 40 on which parts are newly arranged. Data (hereinafter referred to as comparative image data) is compared. Then, the displacement amount observation unit 35 derives a displacement amount of the three-dimensional position of the comparison target component with respect to the three-dimensional position of the comparison source component. The basic image data is derived by the image processing unit 25 by causing the three-dimensional measuring device 20 to capture an image of a state in which the target component is ideally arranged on the kitting tray 40, for example, serving as a layout sample. This is image data including three-dimensional position information (X, Y, Z coordinate values) about the target part. The comparison image data is also image data including the same three-dimensional position information acquired by causing the three-dimensional measuring device 20 to image the kitting tray 40 on which the target part is arranged in the learning process.

The reward setting unit 36 associates the arrangement operation (behavior pattern) executed by the robot hand 10 with the arrangement state of the target component arranged according to the action pattern, and performs a process of giving a reward R to the action pattern. . Specifically, the reward setting unit 36 acquires, from the drive control unit 31, action pattern control data executed by the robot hand 10 when picking and releasing a certain target part. In addition, the reward setting unit 36 acquires the displacement amount data derived by the displacement amount observation unit 35 for the target part arranged according to the behavior pattern. A reward R is given to the behavior pattern based on the behavior pattern control data and the displacement amount data.

The reward R can be set so that a larger value is given as the displacement amount is smaller. For example, when the displacement amount is larger than a preset threshold value Th, the reward R = 0 is set, but when the displacement amount is smaller than the threshold value Th, the reward R> 0 is set. And when the said displacement amount is smaller than threshold value Th, it can set so that the bigger reward R may be given, so that a displacement amount is smaller. In addition, the reward R is given a larger value when, for example, the arrangement where the center of gravity of the parts is lower is not performed, the overlapping of the parts is less, or the tact time of one arrangement cycle is shorter. Can be set to

The value function updating unit 37 updates the value function that defines the value Q (s, a) of the action pattern of the robot hand 10 according to the reward R set by the reward setting unit 36. The value function updating unit 37 updates the value function using an update formula of the value Q (s, a) represented by the following formula (1).

In the above formula (1), “s” represents the state of the robot hand 10 and “a” represents the action of the robot hand 10 according to the action pattern. The state of the robot 2 changes from the state “s” to the state “s ′” by the action “a”. R (s, a) represents the reward R obtained by the transition of the state. The term with “max” is obtained by multiplying the value Q (s ′, a ′) by “γ” when the action “a ′” having the highest value in the state “s ′” is selected. “Γ” is a parameter called an attenuation factor, and is in a range of 0 <γ ≦ 1 (for example, 0.9). “Α” is a parameter called a learning rate, and is in a range of 0 <α ≦ 1 (for example, 0.1).

The above equation (1) is based on the reward R (s, a) set by the reward setting unit 36 for the action “a”, and the value Q (s, a) of the action “a” in the state “s”. Represents an update formula for updating. In other words, the above equation (1) is such that the value Q (s ′, a ′) of the action “a ′” in the state “s ′” is greater than the value Q (s, a) of the action “a” in the state “s”. And the reward R (s, a) is larger, the value Q (s, a) is increased, and on the contrary, the value Q (s, a) is decreased. In other words, the value function updating unit 37 updates the value function using the update formula represented by the above formula (1), thereby the value Q (s, a of the certain action “a” in the certain state “s”. ) To the reward R set for the action “a” and the value Q (s ′, a ′) of the best action “a ′” in the next state “s ′” by the action “a”. I try to get closer.

<Machine learning processing by the learning unit>
9 and 10 are flowcharts showing an example of the learning operation of the layout rule. First, the information acquisition unit 33 of the control unit 30 acquires basic image data of the kitting tray 40 on which parts are arranged and shape data of the kitting tray 40 itself (step S11). The basic image data is captured by the second camera 22 of the three-dimensional measuring device 20 in the kitting tray 40 in which a plurality of types of parts W0 to W32 are ideally arranged as shown in FIG. Acquired based on the three-dimensional image data (ideal information) obtained in this way. The shape data is data relating to the size of the outer frame portion 41 of the kitting tray 40, the size and depth of each inner frame portion 42, and the like. The information acquisition unit 33 acquires the three-dimensional image data obtained by imaging the empty kitting tray 40 with the second camera 22 or the data provided from the input unit 26 as the shape data.

The rule setting unit 32 initially sets a layout rule based on the ideal layout information and the shape data of the kitting tray 40 described above. In other words, it is determined which parts W0 to W32 are arranged in which accommodating parts A1 to A3 of the kitting tray 40. When setting the layout order, basic rules regarding the layout order and layout mode as described above with reference to FIGS. 4 to 8 are stored in the control unit 30 in advance, and are used for initial setting of the layout rules. The This initially set layout rule is corrected according to the learning result (evaluation by the evaluation unit) by the learning unit 34.

Next, the information acquisition unit 33 acquires the shape and storage position information of the target part arranged on the kitting tray 40 (step S12). That is, information regarding the shapes of the components W0 to W32 and position information of the component storage trays for storing the components W0 to W32, respectively, with respect to the robot hand 10 are acquired. These pieces of information can be acquired from the three-dimensional image data based on the imaging result of the first camera 21 of the three-dimensional measurement apparatus 20 or the input data given from the input unit 26. The above steps S11 and S12 are preparations for the learning process.

Upon entering the learning process, the information acquisition unit 33 acquires the position information of the kitting tray 40 from which parts will be arranged from the imaging result of the second camera 22 (step S13). That is, the position information of the kitting tray 40 with respect to the robot hand 10 is acquired. Subsequently, the information acquisition unit 33 causes the first camera 21 to image the component storage tray that stores the layout target component, and obtains the three-dimensional position information of the layout target component based on the result of the object recognition process of the image processing unit 25. Obtain (step S14). Thereby, the coordinate value in the said component storage tray of the target component arranged from now on is acquired. The position information acquired by the information acquisition unit 33 is given to the drive control unit 31 through the rule setting unit 32.

The drive control unit 31 operates the robot hand 10 based on the layout rule set by the rule setting unit 32 and the position information acquired by the information acquisition unit 33, and sequentially picks the target parts (step S15). Then, based on the imaging result of a component recognition camera (not shown) that captures the part gripped by the hand unit 16 of the robot hand 10 from the lower surface side, the control unit 30 determines whether or not the target component is gripped by the hand unit 16. Determination is made (step S16). If the target part is not gripped, that is, if the hand unit 16 has failed to grip the part (NO in step S16), the process returns to step S14 to retry the picking of the target part.

When the target part is gripped (YES in step S16), the drive control unit 31 drives the robot hand 10 to transport the picked target part to the kitting tray 40, and also uses the layout rule and the position information. Based on this, the target part is released at a predetermined XYZ position (step S17). Thereby, the arrangement of one target part is completed.

Thereafter, it is confirmed whether or not all the scheduled part arrangements have been completed (step S18). If all arrangements have not been completed (NO in step S18), it is subsequently confirmed whether or not the success rate of gripping the parts by the hand unit 16 is good (step S19). If the gripping success rate is good (YES in step S19), it can be said that good picking is being performed based on the object recognition processing result acquired in step S14. In this case, the process proceeds to step S15, and the drive control unit 31 executes the picking of the next target part. On the other hand, if the gripping success rate is not good (NO in step S19), it can be said that there is a difference between the object recognition processing result and the actual picking. In this case, the process returns to step S14, and the first camera 21 is again imaged of the component storage tray, and the object recognition process is performed.

When all arrangements are completed (YES in step S18), the process proceeds to the flow of FIG. 10, and information on the kitting tray 40 after the arrangement is completed is acquired (step S21). Specifically, the imaging control unit 24 of the three-dimensional measuring device 20 causes the second camera 22 to image the kitting tray 40 after the arrangement, and the image processing unit 25 indicates a three-dimensional position indicating the arrangement position of the components on the kitting tray 40. Deriving information. The displacement amount observation unit 35 of the learning unit 34 acquires image data including the three-dimensional position information of such a component from the camera control unit 23 as the above-described comparison image data. It is unclear how the parts behave in the kitting tray 40 after releasing the parts gripped by the hand unit 16. The execution result of one action pattern of the robot hand 10 that grasps and picks one part at a specific position and releases it at a specific height position is grasped in this step S21.

The displacement amount observation unit 35 compares the comparison image data acquired in step S21 with the basic image data acquired in step S11, and the three-dimensional position of the component in the basic image data of the three-dimensional position of the component in the comparison image data. The amount of displacement with respect to is derived (step S22). In addition, the image data of the kitting tray 40 after the arrangement obtained in the learning process so far can be used as the basic image data of the comparison source. In this case, the positional stability of the components in the kitting tray 40 after the arrangement can be evaluated. Further, it is desirable that the displacement amount is mainly a displacement amount of the center of gravity position of the component.

Subsequently, the reward setting unit 36 determines whether or not the displacement amount is larger than a preset threshold value Th (step S23). The large amount of displacement means that in the action pattern of the robot hand 10 this time, the component is not arranged at the intended position, or the component cannot be arranged stably. When the displacement amount is equal to or greater than the threshold Th (YES in step S23), the reward setting unit 36 gives a reward R of “0; zero” to the action pattern of the robot hand 10 (step S24). On the other hand, when the amount of displacement is less than the threshold Th (NO in step S23), the reward setting unit 36 gives a reward R greater than “0; zero” to the action pattern of the robot hand 10. (Step S25).

Thereafter, the value function updating unit 37 updates the value function that defines the value Q (s, a) of the action pattern of the robot hand 10 by using the update formula of the above formula (1) (step S26). Each process shown in steps S13 to S26 is a process executed in one cycle of the learning process by the learning unit 34. The learning unit 34 determines whether or not the number of learning has reached the predetermined number N (step S27). If the predetermined number N has not been reached (NO in step S27), the learning unit 34 returns to step S13, causes the parts to be arranged on the next kitting tray 40, and repeats the learning process. On the other hand, if the predetermined number N has been reached (YES in step S27), the learning unit 34 ends the learning process.

<Modification of machine learning process>
In the example of the machine learning process described above, an example is shown in which the information acquisition unit 33 acquires basic image data (ideal layout information) of the kitting tray 40 on which the components are arranged in step S11 of FIG. Instead of this, the ideal layout information may not be given to the information acquisition unit 33, and only the information related to the kitting tray 40 on which no component is arranged and its storage unit, and the information related to the size of the component may be acquired.

That is, in this modification, in step S11 of FIG. 9, the information acquisition unit 33 acquires only the shape data of the kitting tray 40 itself. Then, the rule setting unit 32 initially sets a provisional layout rule based on the shape data of the target part acquired in step S12. This layout rule is modified according to the learning result of the learning unit 34 (evaluation by the evaluation unit).

According to this modification, it is possible to detect a layout rule close to the ideal layout by machine learning even if the ideal layout information is not given. In the ideal catering information, the catering rules tend to be set according to a fixed premise. For example, a layout rule tends to be set based on the premise that a small-sized component is distributed in a small-sized storage section prepared in the kitting tray 40. However, based on such premise, there are cases where good parts cannot be arranged. This point will be described with reference to FIG.

FIGS. 11A and 11B are diagrams showing examples of arrangement of the large-sized component W61 having a relatively large size and the small-sized component W62 having a relatively small size on the kitting tray 40. FIG. The kitting tray 40 includes a wide accommodating portion A16 having a relatively large accommodating space and a narrow accommodating portion A17 having a relatively narrow accommodating space.

FIG. 11A shows a state in which a large size part W61 is arranged in the wide accommodation part A16 and a small size part W62 is arranged in the narrow accommodation part A17. The arrangement example shown in FIG. 11A is in line with a general arrangement concept in which a small-sized component is arranged in a small-sized container. However, in the actual layout result, since the number of small size parts W62 is large, the small size parts W62 are piled up in the narrow housing portion A17. That is, the position of the center of gravity of many small-sized components W62 is located higher than the bottom plate 43 in the narrow housing portion A17. In this case, it is not preferable because the small-sized component W62 is likely to overflow from the kitting tray 40 (narrow accommodating portion A17).

On the other hand, in the arrangement example shown in FIG. 11 (B), a large size part W61 is arranged in the narrow accommodating part A17, and a small size part W62 having a large number of parts is arranged in the wide accommodating part A16. In this arrangement example, the positions of the center of gravity of the small-sized component W62 and the large-sized component W61 are located lower than the bottom plate 43 in the respective accommodating portions A16 and A17. Thereby, the arrangement | positioning in which components W61 and W62 cannot fall easily from the kitting tray 40 is realizable. It is possible to detect the layout mode as shown in FIG. 11B by not giving the ideal layout information and learning the layout mode that lowers the center of gravity as much as possible by the learning process by the learning unit 34.

[Inventions Included in the Embodiments]
The specific embodiments described above mainly include inventions having the following configurations.

A component arrangement device for a kitting tray according to one aspect of the present invention is a component arrangement device that arranges a plurality of types of components of different sizes on a kitting tray having a plurality of storage units, and picking and releasing the components. An arrangement that has a possible head part, picks and conveys the target part from the plurality of types of parts at the storage position of the part, and releases the target part from the head part to the kitting tray. A robot hand that performs an operation; and a control unit that controls the operation of the robot hand, wherein the control unit is configured to control the kitting tray of the component according to the types of the plurality of components and the plurality of storage units. A rule setting unit for setting a distribution rule to the robot, and causing the robot hand to execute the distribution operation based on the distribution rule Comprising a drive control unit.

According to this component arrangement apparatus, the arrangement rule is set according to the aspect of the plurality of accommodating parts of the kitting tray and the aspect of the parts arranged on the kitting tray. That is, instead of strictly determining the arrangement procedure by the robot hand by programming or the like, the rule setting unit can flexibly determine the arrangement rule according to the aspect of the part or the accommodation unit. Therefore, it is possible to properly distribute the components to the kitting tray without performing programming or the like that requires time and skill.

In the above-described component arrangement apparatus, the rule setting unit sets an arrangement rule that arranges components in which the arrangement position is lower when the arrangement in which the components overlap vertically in the kitting tray is performed. It is desirable.

According to this component arrangement apparatus, it is possible to prevent a problem that a component whose arrangement position should be lower is arranged on a component whose arrangement position is upper.

In the above-described component arrangement device, the rule setting unit is more effective than the arrangement in which the components overlap in the vertical direction when both the arrangement in which the components are arranged in the horizontal direction and the arrangement in the vertical direction are executable in the kitting tray. It is desirable to set a layout rule that gives priority to the layout in which the parts are arranged in the horizontal direction.

According to this component arrangement apparatus, it is possible to perform an arrangement that effectively uses the horizontal space of the kitting tray without causing an arrangement in which components are stacked in the vertical direction.

In the above-described component arrangement device, the rule setting unit sets an arrangement rule for arranging components with a small occupation area in plan view in advance when performing arrangements in which the components overlap in the vertical direction in the kitting tray. It is desirable to do.

According to this component arrangement apparatus, a large component having a large occupied area blocks the housing portion of the kitting tray, and a small component having a small occupied area cannot be arranged, or a small component is arranged on the large component. Can be prevented.

In the above-described component arrangement apparatus, the rule setting unit includes a first arrangement state in which the height position of the center of gravity of the component is the first position, and a height of the center of gravity of the component as the arrangement state of the component on the kitting tray. It is desirable to set the layout rule so that the component is in the first layout state when the position can be in the second layout state in which the position is a second position higher than the first position.

According to this component arrangement apparatus, each component can be arranged in the kitting tray with a lower center of gravity. Thereby, when moving a kitting tray after arrangement | positioning, components can be made hard to fall from a kitting tray.

In the above-described component arrangement apparatus, an imaging device that acquires a three-dimensional image of the kitting tray after the arrangement operation by the robot hand is performed, and a component arrangement state in the kitting tray based on the three-dimensional image It is preferable that the rule setting unit further sets the layout rule according to the evaluation of the evaluation unit.

According to this component layout apparatus, it is possible to cause the rule setting section to perform machine learning on a technique for performing better layout, for example, a picking mode and a release position by the head section through the evaluation of the evaluation section. And since the said rule setting part can set a layout rule based on the said machine learning, the layout rule excellent in the layout property without programming can be set.

In the component arrangement apparatus, the rule setting unit acquires ideal arrangement information in a state where the plurality of types of components are ideally arranged with respect to the kitting tray, and the arrangement is based on the ideal arrangement information. It is desirable to initialize a rule and modify the serving rule according to the evaluation of the evaluation unit.

According to this component layout apparatus, the layout rule that is initially set based on the virtual layout information can be modified according to the machine learning of the layout method by the rule setting unit, and the optimal layout rule can be set ultimately. .

In the component arrangement apparatus, the rule setting unit acquires information about the kitting tray and its storage unit and information about the size of the component, and initializes the layout rule based on the information, and the evaluation unit It is desirable to modify the serving rule in accordance with the evaluation.

According to this component arrangement apparatus, an arrangement rule is initially set based on information relating to the kitting tray and the size of the component, and the arrangement rule is corrected according to the machine learning of the arrangement method by the rule setting unit. Therefore, even if the ideal layout information as described above is not given, it is possible to detect a layout rule close to the ideal layout by machine learning.

As described above, according to the present invention, it is possible to provide a component arrangement device for a kitting tray that can accurately arrange various components on the kitting tray using a robot hand.

[Explanation of symbols]
Wa, Wb, Wc, Wd, W1 to W62 Parts (target parts)
Ta, Tb Component storage tray (storage position)
A1 to A3, A11 to A17 Accommodating portion G1, G2 Center of gravity h1, h2 First position, second position 1 Component arrangement device 10 Robot hand 15 Head unit 16 Hand unit 20 Three-dimensional measuring device (imaging device)
DESCRIPTION OF SYMBOLS 21 1st camera 22 2nd camera 23 Camera control part 24 Imaging control part 25 Image processing part 30 Control part 31 Drive control part 32 Rule setting part 33 Information acquisition part 34 Learning part (evaluation part)
35 Displacement observation unit 36 Reward setting unit 37 Value function update unit 40 Kitting tray 41 Outer frame unit 41
42 Inner frame part 42
42T Upper edge 42T
43 Bottom plate 43

Claims

A component arrangement device that arranges a plurality of types of parts having different sizes on a kitting tray having a plurality of storage units,
A head portion capable of picking and releasing a component, picking and transporting the target component from the plurality of types of components at the storage position of the component, and transporting the target component from the head portion; A robot hand that performs a catering action to be released to the kitting tray;
A control unit for controlling the operation of the robot hand,
The controller is
A rule setting unit that sets a layout rule of the parts to the kitting tray according to the types of the parts and the housing parts,
A component arrangement device for a kitting tray, comprising: a drive control unit that causes the robot hand to execute the arrangement operation based on the arrangement rule.
In the component arrangement apparatus to the kitting tray of Claim 1,
The rule setting unit sets a layout rule for first arranging components whose layout position is lower when performing layout that causes vertical overlap between components in the kitting tray. apparatus.
In the component arrangement apparatus to the kitting tray of Claim 1,
In the kit setting tray, when both the arrangement in which the parts are arranged in the horizontal direction and the arrangement in which the parts overlap in the vertical direction can be executed, the arrangement in which the parts are arranged in the horizontal direction rather than the arrangement in which the parts overlap in the vertical direction is possible. A parts distribution device for kitting trays that sets the distribution rules to be executed with priority.
In the component arrangement apparatus to the kitting tray of Claim 2,
The rule setting unit sets a layout rule that sets a layout rule that allows a component with a small occupation area to be arranged in plan view when performing layout in which the components overlap in the vertical direction between the components in the kitting tray. Catering equipment.
The component arrangement device for the kitting tray according to any one of claims 1 to 4,
The rule setting unit includes a first arrangement state in which a height position of the center of gravity of the part is the first position, and a height position of the center of gravity of the part is the first position. A component arrangement device for a kitting tray, wherein the arrangement rule is set so that the component takes the first arrangement state when the second arrangement state, which is a higher second position, can be taken.
In the component arrangement device for the kitting tray according to any one of claims 1 to 5,
An imaging device that acquires a three-dimensional image of the kitting tray after the catering operation by the robot hand is executed;
An evaluation unit that evaluates the arrangement state of components in the kitting tray based on the three-dimensional image, and
The rule setting unit is a component layout device for a kitting tray that sets the layout rule according to the evaluation of the evaluation unit.
In the component arrangement apparatus to the kitting tray of Claim 6,
The rule setting unit acquires ideal layout information in a state where the plurality of types of components are ideally arranged with respect to the kitting tray, and initially sets the layout rule based on the ideal layout information. A parts layout device for a kitting tray that corrects the layout rules according to evaluation by an evaluation unit.
In the component arrangement apparatus to the kitting tray of Claim 6,
The rule setting unit acquires information related to the kitting tray and its storage unit and information related to the size of a part, initializes the layout rule based on the information, and determines the layout according to the evaluation of the evaluation unit. Parts arrangement device for kitting tray to correct rules.