JP6811505B1 - Picking device and picking method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a picking technique which is highly versatile, can be introduced at low cost, and can be operated promptly. SOLUTION: The picking device 1 picks up the articles one by one from the sky from a container C in which articles of the same shape are randomly stored, and obtains a visual sensor 10 for acquiring three-dimensional information of the article group. Based on the three-dimensional information of the article group, the collation unit 33 forcibly collates the articles with one of the sphere, ellipsoid, cube, rectangular parallelepiped, and other sample figures to calculate the center position of each. The tactile sensor 27 is provided at the tip of each of the tactile sensors 27, and the gripping unit 21 consisting of at least two arms 25 for gripping the article by narrowing the interval and the gripping unit 21 are lowered from the sky at any center position. 1. A picking device 1 including an open / close control unit 38 that controls an operation of gripping and ascending an article by narrowing the distance between the arms 25 based on the detection of contact with the article. [Selection diagram] Fig. 1

Description

The present invention relates to a picking device (pickup device) used when placing parts bulk in a container on a conveyor or the like, and more particularly to a technique for realizing simple picking based on visual information and tactile information.

Conventionally, there is known a picking robot that takes out a large number of parts of the same shape that are piled up in bulk in a container and places them on a belt conveyor or the like. Here, since the postures of the parts are not uniform and are stacked in the depth direction, the conventional picking robot inputs accurate shape data of the parts and stereo-images a group of articles by a high-precision visual sensor. , Data collation was performed by contour matching and point cloud matching to determine the gripping position of the article. In other words, all the calculations were completed before moving to the gripping motion, and although the motion was different each time, the approach to the article and the motion during gripping were individually determined in advance.
Therefore, it is necessary to use a high-performance model for the visual sensor, such as acquiring high-definition information using a plurality of optical systems, and there is a problem that the introduction cost becomes extremely high.
In addition, although CAD data is generally adopted as the accurate shape data of the article, it is necessary to input the data individually every time the target article changes, and this and a large amount of information from the visual sensor are input. In order to collate, a high-performance processing system (PC, etc.) is required, and there is a problem that the cost is high in any case.
In addition, it is necessary to tune each time the type of parts is changed to determine the gripping position, and there is a problem that it may take several months or more to operate.
That is, the conventional technique has a problem that high-definition processing is performed in advance to perform picking, and fine tuning is required separately, the system becomes extremely expensive, and it takes time to operate.

Japanese Unexamined Patent Publication No. 11-123681 Japanese Unexamined Patent Publication No. 2010-184308 Japanese Unexamined Patent Publication No. 2011-000685 Japanese Unexamined Patent Publication No. 2011-183537 Japanese Unexamined Patent Publication No. 2013-21250

The present invention has been made in view of the above, and an object of the present invention is to provide a picking technique that is highly versatile, can be introduced at low cost, and can be operated quickly.

The picking device according to claim 1 is a picking device that picks up articles of the same shape one by one from the sky from a container in which articles of the same shape are randomly stored, and acquires three-dimensional information of a group of articles in the container. Forcing an article by one of a circle, a sphere, an ellipsoid, a square, a rectangle, a cube, a rectangular parallelepiped, or another sample figure based on the visual sensor to be used and the three-dimensional information of the article group acquired by the visual sensor. One of a collation calculation means for collating and calculating the center position of each, a grip unit consisting of at least two arms having a tactile sensor at the tip and gripping an article by narrowing the interval, and a grip unit. It is characterized by being provided with a gripping control means for controlling the movement of the article by lowering it from the sky at the center position and narrowing the arm spacing based on the detection of contact with the article by the tactile sensor.

That is, the invention according to claim 1 eliminates the need for a high-spec sensor or processing system by combining a rough position grasp of bulk articles by forced collation and a position grasp based on on-site correction by a tactile sensor. As a result, it is possible to construct a system that is highly versatile and can be introduced at low cost. Since on-site correction or on-site response is performed, time-consuming tuning is not required, and quick operation can be realized.

The operation of picking up from the sky is generally up and down on the vertical line, but it does not prevent the approach and pickup modes at an angle.
The three-dimensional information refers to the position information on the surface of the article group. That is, it is information including depth (depth) information. It does not have to be as dense as high-resolution image information, but may be three-dimensional information that is appropriately meshed or dotted so that the shape of the article can be roughly grasped. This reduces the number of data and eliminates the need for high specifications for visual sensors and processing systems. The visual sensor is sometimes called a depth sensor. Further, the measurement method of the three-dimensional information is not limited, and examples thereof include an Active Stereo method, a Time of Flight method, and a Code Light method.
The sample figure may be set by the user, or the optimum candidate may be automatically selected based on the three-dimensional information from the visual sensor. A simple sample figure can be adopted and the processing load can be reduced. A large-scale point cloud processing library such as PointCloudLibry (PCL) may be used for the processing and collation processing of the three-dimensional information. Prior to collation, it is also possible to perform preprocessing for enlarging / reducing the set sample figure so as to be closer to the article. In collation, the determination may be made based on the fact that the overlap between the figure and the contour of the target is the largest, and an example of determining by comparing the degree of coincidence with the normal direction of each surface can be given. In addition to rotating the figure for trial evaluation, the size of the set figure may be adjusted separately from the preprocessing based on the information in the depth direction.
The center position calculated by collation is the center of the figure or the center of gravity, and depending on the article, the center position may be shifted according to a predetermined rule.
The fact that it is composed of at least two arms means that in the case of two arms, the tactile sensors may face each other so as to face each other, and in addition, gripping by three arms and gripping by four arms are not hindered. If there are three, it is preferably 120 ° rotationally symmetric, and if there are four, it is preferably 90 ° rotationally symmetric.
The tactile sensor is a sensor that detects the presence or absence of contact with an object and outputs the degree of contact as an appropriate numerical value. The so-called pressure sensor outputs one numerical value or one numerical value as an average value, whereas the tactile sensor outputs, for example, pressure values at a plurality of points in a plane substantially perpendicular to the direction in which the load is applied, and makes contact with each other. It is a sensor that can evaluate the degree or appearance of (pinpoint pressure, tilted contact to the right, pulsating, etc.) (in this sense, the pressure sensor is within a range close to some extent. It can also be used as a tactile sensor by combining a plurality of them and using each output value).
It is preferable that the tactile sensor portion itself comes into surface contact with the article and picks it up. This is because the gripping pressure can be monitored in real time to detect that the article is about to fall and the gripping pressure can be increased. The surface is preferably made of a material having a large frictional force such as rubber.
The pick-up order is preferably the height order of the articles based on the three-dimensional data.
Gripping an article based on the detection of contact with an article by a tactile sensor does not only mean that the article is narrowed as it is after the contact, but may also include motion control such that the gripping unit is further lowered by 1 mm and then narrowed. It shall be.
The mode of narrowing the arm spacing is a mode in which the tip is narrowed by narrowing the base end spacing of the arm even if the rotation is such that the tip spacing is narrowed around each base end side such as the mounting position of the arm. It may be (in this case, the arms are preferably parallel to each other), or it may be a combination of these movements.

The picking device according to claim 2 is the picking device according to claim 1, wherein the tactile sensor is attached to the mounting surface at the tip of the arm as a flat pad body that grips the article by surface contact. The normal direction is 75 ° to 55 ° with respect to the line connecting the base end of the arm and the mounting surface.

That is, the invention according to claim 2 has a structure in which the tip of the gripping unit is widened to alleviate entry mistakes and realize stable gripping at a deep position.

Needless to say, the pad bodies have an opposing positional relationship (when there are three or more arms, the pad bodies are symmetrically arranged inward with respect to the descending direction).
Note that 75 ° to 55 ° means that the tip of the arm is open at an angle of 15 ° to 35 ° with respect to the arm portion up to that point. If it exceeds 75 °, the effect of making an angle is not sufficiently exhibited, and if it is less than 55 °, it spreads too much and the gripping stability is not exhibited.

The picking device according to claim 3 is the picking device according to claim 1 or 2, in which one of the tactile sensors detects contact with the article, and then the tactile sensor of the other arm contacts the article. The gripping control means is provided with a movement control means for moving the gripping unit to the other arm side until the detection is detected, and a correction means for correcting the center position based on the movement distance of the gripping unit by the movement control means. It is characterized in that the motion control for gripping the article is performed from the center position corrected by.

That is, the invention according to claim 3 realizes more accurate position grasping based on on-site correction by a tactile sensor.

The motion control for gripping the article from the corrected center position may be a mode in which the gripping unit is horizontally moved to narrow the distance between the arms as it is from the new center position, and after reaching the new center position again. It descends (it may be ascended and then lowered depending on the aspect of the specification), and after contact detection, the second centering (and the third centering ...) is performed before the pinching is started. May be good. Correction can be rephrased as correction.
When the gripping unit is moved to the other arm side until the tactile sensor of the other arm detects the contact with the article, the tactile sensor that first detects the contact may keep its posture ( In this case, the movement of the gripping unit causes the article to move away from the article), and the movement of the gripping unit may be controlled to maintain contact.

The picking device according to claim 4 is the picking device according to claim 1, 2 or 3, wherein the tactile sensor is a displacement value or a displacement value derived from contact at at least two places, a front side and a rear side in the descending direction. Each pressure value is output, the output value from the tactile sensor is input, the output value on the front side is above a certain value, and the output value on the rear side asymptotically approaches the output value on the front side within a predetermined range. The gripping control means narrows the arm spacing, and when it is determined by the determining means that the gripping control means does not fit, the arm spacing is widened and the gripping unit is lowered again to determine whether the gripping unit is fitted or not yet fitted. The operation of narrowing the arm spacing is repeated until it is determined by the determination means that the arm spacing is narrowed, and when it is determined that the arm spacing is within the determination means, the operation control of raising the gripping unit is performed.

That is, the invention according to claim 4 performs a so-called search operation, advances the arm to a certain depth, and realizes stable surface contact gripping. In the aspect subordinate to claim 3, more stable and error-free pickup is realized.

The front side and the rear side in the descending direction can also be expressed as the front side and the base side in the extending direction of the arm.
The tactile sensor may output data at four locations, not only at two locations in the front and rear, but also at two locations on the left and right. Further, since the tactile sensor is located at the tip of each arm, the determination may be made using the output values of the two front and rear positions of each tactile sensor. When it is lightly pinched, the bias of each tactile sensor may be confirmed, and if it is not even, the center position may be reset.
The displacement value depends on the configuration of the tactile sensor, but may be, for example, the amount of sinking of the pad, or the amount of displacement converted based on the transmittance or the transmittance of light in the pad may be adopted.
The amount of expansion and the amount of descent can be set as appropriate, but an example of setting each to several millimeters can be given. In the case of two arms, the front side of both sensors shifts the center to the position where the contact detection limit is reached almost at the same time, and the correction or drive, which can be called a trace approach, is performed by descending 1 mm again to make a judgment. It can also be done (the same applies to three or more arms).

The picking method according to claim 5 is a gripping unit in which arms arranged symmetrically with a tactile sensor at the tip grip and lift an article to be gripped evenly at close intervals, and three-dimensional information of an article group. This is a picking method in which articles of the same shape are picked up one by one from the sky from a container in which articles of the same shape are randomly stored, using a visual sensor to acquire a predetermined ellipsoidal shape, rectangular parallelepiped shape, etc. By applying the sample figure of the above to the three-dimensional information of the article group acquired by the visual sensor, the articles are forcibly collated to calculate the center position of each, the article to be lifted is determined, and the sky above the center position of the article. When one of the tactile sensors detects contact with the article, the gripping unit is moved to the other arm side until the tactile sensor of the other arm detects the contact with the article. The center position is corrected based on the moving distance of the gripping unit, and the article is gripped and lifted from the vertical line of the corrected center position.

That is, the invention according to claim 5 eliminates the need for a high-spec sensor or processing system by combining a rough position grasp of bulk articles by forced collation and an accurate position grasp based on on-site correction by a tactile sensor. As a result, it is possible to construct a system that is highly versatile and can be introduced at low cost. Since on-site correction is performed, time-consuming tuning is not required, and quick operation can be realized.

The picking method according to claim 6 is the picking method according to claim 5, wherein the tactile sensor outputs contact pressure at at least two locations, a front side and a rear side in the descending direction, and the front side. After repeating the operation of opening the arm, lowering the gripping unit, and closing the arm until the contact pressure on the rear side is asymptotically close to the contact pressure on the front side and falls within the predetermined range. , It is characterized by lifting an article.

That is, the invention according to claim 6 performs a so-called search operation, advances the arm to a certain depth, and realizes a stable and error-free pickup.

Contact pressure refers to the displacement value or pressure value derived from contact.

According to the present invention, it is possible to provide a picking technique that is highly versatile, can be introduced at low cost, and can be operated quickly.

It is a schematic diagram of the appearance of the picking apparatus 1. This is an example of a Stanford bunny with a height of about 77 mm and a reconstructed image of three-dimensional information obtained by photographing it with a visual sensor. It is a schematic diagram which showed the appearance structure of the gripping unit 21. It is explanatory drawing which showed the structure of the tactile sensor 27. It is a block diagram which showed the functional structure of the processing apparatus 30. It is a photograph showing an image of a Stanford bunny in a container taken by a visual sensor and an example of forcibly collating it in a spherical shape using the three-dimensional information. It is a figure which showed the analysis result of the sensor displacement amount of the pick-up operation by a person. It is a flowchart which shows the flow of a picking process.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Here, an example of picking up bulk parts by two arms will be described.

FIG. 1 is a schematic external view of the picking device of the present invention. For convenience of explanation, the scale of each configuration is not necessarily the same in each figure.
The picking device 1 is a device that picks up articles of the same shape that are bulk-loaded and stored in the container C one by one and places them on a belt conveyor (not shown). In the following, the details up to the point of picking up will be described in detail, and then the description of how to place the conveyor on the conveyor will be omitted. However, it is easy to align and place them using general-purpose technology.

In the conventional picking device, the CAD data of the article is input in advance, and the position and posture of the article are completely collated based on the three-dimensional information (three-dimensional data) or depth data taken in high definition and high resolution. After grasping all of the above, the gripping operation is precisely determined and the article is picked up.
On the other hand, the present invention is largely different in that the position and posture of the article are roughly grasped and the gripping operation is determined in response to the site, as will be described below.

The picking device 1 is mainly composed of a visual sensor 10, an articulated robot 20, and a processing device 30.

The visual sensor 10 photographs an area including the container C from the sky above the mounting surface of the container C, and acquires three-dimensional information of the article group. More specifically, the unevenness information of the article group including how the articles are overlapped is acquired. The shooting method is not particularly limited, but here, the Code Light (IR) method is adopted. An example of such a visual sensor is Intel RealSense® SR300. Equipped with an infrared projector, an infrared camera, and an RGB camera, it is inexpensive and can easily acquire three-dimensional information.
FIG. 2 is an example of a Stanford bunny having a height of about 77 mm and a reconstructed image of three-dimensional information obtained by photographing the Stanford bunny with a visual sensor.
The dot spacing of the reconstructed image can be arbitrarily set below the resolution, but can be 1 mm to 5 mm depending on the size and shape of the parts to be handled. In other words, since the article is reconstructed with hundreds to thousands of dots and collated with the collation figure described later, neither the visual sensor 10 nor the processing device 30 needs to have high specifications and can be introduced at low cost.

The articulated robot 20 is composed of a gripping unit 21 and an articulated portion 22 that freely moves and orients the gripping unit 21. The articulated portion 22 allows the gripping unit 21 to perform an approach-reverse operation for picking up, and after picking up, a transfer operation and a placement operation on a conveyor. In the present embodiment, the approaching motion and the retracting motion are performed on the vertical line, and the stretching direction (orientation direction) of the gripping unit 21 is also on the vertical line.
The gripping unit 21 is composed of two arms 25 and a base end driving unit 26 for attaching and opening and closing the arms 25. FIG. 3 is a schematic view showing the external configuration of the gripping unit 21.
The arms 25 are plates of 25 mm × 150 mm × 6 mm, facing each other, and the tip 30 mm portion is spread outward at an angle of 25 ° with the previous portion. By expanding the diameter of the arm 25 to the outside in this way, stable gripping is achieved, just as when a person picks a thing with the index finger and the thumb, the pad of the finger is tilted and a suitable knob is realized. Realize.
Further, a tactile sensor 27, which is a cylindrical pad (diameter 18 mm, height 7 mm) having a flat outer shape, is attached to the inside of the tip portion. The surface material is rubber, and the article is gripped by surface contact on the pad surface.
The tactile sensor 27 knows how to apply the load at four locations inside the peripheral edge: 0 o'clock, 3 o'clock, 6 o'clock, and 9 o'clock. FIG. 4 is an explanatory diagram showing the structure of the tactile sensor 27. 4a is an external configuration, FIG. 4b is a schematic plan view showing the internal structure of the sensor, and FIG. 4c is an explanatory view showing a state of light transmission when a load is applied.
As shown in the figure, the inside of the pad is made of sponge, and the sponge density that increases according to the load on the pad is quantified as a change in the amount of light transmitted inside, converted into a pressure value, and output. By doing this at four locations, it is possible to understand how the load is applied.
In particular, the tactile sensor 27 is attached to the arm 25 with the 0 o'clock direction aligned with the tip end side (front side) of the arm 25 and the 9 o'clock direction with the proximal end driving unit 26 side (rear side). As will be described later, sufficient gripping is realized by knowing how the load is applied (the degree of surface contact with the article) at at least two places in the front and rear.
As an example of the tactile sensor 27 described above, the shock pot POTU-001-1 manufactured by Touchence Co., Ltd. can be mentioned and can be introduced at low cost.

Next, the processing apparatus 30 will be described.
Since the processing device 30 can use a so-called personal computer and general-purpose products such as a CPU, memory, storage, and monitor can be adopted, detailed description of hardware will be omitted here.
Specific functional configurations of the processing device 30 include a three-dimensional information input unit 31, a sample graphic setting unit 32, a collation unit 33, a center calculation / correction unit 34, an articulated control unit 35, and tactile information. It has an input unit 36, a determination unit 37, and an open / close control unit 38. FIG. 5 is a block diagram showing a functional configuration of the processing device 30. The articulated control unit 35 and the open / close control unit 38 mainly carry out the gripping control according to the claim. These functional units are realized by cooperating with software such as a processing program and an OS installed in the processing device 30 and each hardware.

The three-dimensional information input unit 31 sequentially inputs the three-dimensional information of the article group acquired by the visual sensor 10. It also controls storage in RAM or hard disk as needed.
The sample figure setting unit 32 sets a matching figure based on a simple figure such as a circle, a sphere, an ellipsoid, a square, a rectangle, a cube, or a rectangular parallelepiped, and three-dimensional information input from the three-dimensional information input unit 31. ..
The surface unevenness of the article group can be understood from the three-dimensional information, and the contour of each article in the upper layer can be roughly determined based on the information of the portion where the value in the depth direction has a large difference (especially the contour at a high position). It is preferable to use it). Based on that information, a collation figure is selected. In the present embodiment, collation is performed by a spherical shape. Further, the radius of the sphere is used for collation after being appropriately adjusted. In addition, a spheroid whose three-axis length is adjusted can be used, or a rectangular parallelepiped whose three-sided length is adjusted can be used.

The collation unit 33 forcibly matches the spherical shape, which is a collation figure, with the article based on the three-dimensional information. Since the articles in the upper layer have almost no breaks in the contour, the ones with the smallest break rate are collated (and picked up). Even if the thin part of the article does not have a large step between it and the lower part, other articles may have a random orientation and a large step may occur in the part. Such data can be used to supplement the thin part. The disconnection part can be complemented in the same way.
The collation method is not particularly limited, but the maximum likelihood estimation, for example, the volume overlap ratio by moving the sphere little by little based on the feature quantity such as the surface shape of the target article including the contour and the normal direction of each point. An example can be given in which the position where is highest is calculated, and if it is equal to or more than a predetermined value, it is determined that they are collated (both figures match).
Since the three-dimensional information is input in sequence, the article in the back becomes the three-dimensional information displayed after the pickup, and this updated three-dimensional information is used in the next collation.
FIG. 6 is a photograph showing an image of a Stanford bunny in a container taken by a visual sensor (FIG. 6a) and an example of forcibly collating the Stanford bunny in a spherical shape using the three-dimensional information (FIG. 6b). ..

The center calculation / correction unit 34 sets the center coordinates of the sphere to the center of the article based on the sphere collated by the collation unit 33.
Further, the center calculation / correction unit 34 corrects the center of the article from the moving distance of the gripping unit 21 based on the first contact and the second contact described later.

The articulated control unit 35 controls the articulated unit 22 and sets the gripping unit 21 vertically above the coordinate values of the center coordinates with respect to the article to be gripped, which is set or reset by the center calculation / correction unit 34. Move. At this time, the posture of the gripping unit 21 is controlled so that its center line faces vertically downward.
Then, the articulated control unit 35 lowers the gripping unit 21, grips the article, and then raises (picks up) the article again. After that, the article is appropriately placed on a conveyor or the like, returned to the sky above the container C, and picking is repeated.
Further, when the determination unit 37 detects the first contact (first contact), the articulated control unit 35 moves the gripping unit 21 to the other arm 25 side (the determination unit 37 makes the second contact (second contact). ) Is detected).
Further, when the multi-joint control unit 35 determines that the determination unit 37 is within the predetermined range (when the side bottom portion 37 determines that the arm 25 has entered a deep and stable position and is gripping the article), the multi-joint control unit 35 grips The operation control for raising the unit 21 is performed.

The tactile information input unit 36 sequentially inputs pressure values from a total of eight locations of the two tactile sensors 27.

The determination unit 37 determines whether or not one of the tactile sensors 27 has detected contact with the article as the gripping unit 21 is lowered under the control of the articulated control unit 35 (determines the first contact).
Further, the determination unit 37 determines whether or not the tactile sensor 27 of the arm 25 on the opposite side of the articulated control unit 35 has detected contact with the article (determines the second contact).
Further, the determination unit 37 determines that the pressure value on the front side is equal to or higher than a certain value and the pressure value on the rear side and the pressure value on the front side are determined based on the transition of the pressure value from the front side and the rear side of the tactile sensor 27. It is determined whether the difference between the two is within the predetermined range or not yet (by repeating the process sequentially, it is determined whether the pressure value on the rear side asymptotically approaches the pressure value on the front side and is within the predetermined range).

This determination is based on a prior experiment in which the inventor of the present application attaches the tactile sensor 27 to the abdomen of the index finger and the thumb and grips the article. FIG. 7 is a diagram showing the analysis result of the sensor displacement amount of the pick-up operation by a person. As shown in the figure, when the finger hits the article and then picks it up, the action of deeply inserting the finger and the action of spreading it are repeated as the first exploration motion (in the figure, it appears as three peaks of the displacement amount on the front side). Go deep and pick it up as a stable pinch (in the figure, the amount of displacement on the rear side is about the same as the amount of displacement on the front side). In the picking device 1, this operation is performed as a field response.

The open / close control unit 38 performs an operation of narrowing (finally holding an article) or widening the distance between the arms 25 symmetrically with respect to the center line of the gripping unit 21.
In particular, in cooperation with the determination unit 37, the article enters a certain deep position to perform stable gripping. That is, when the opening / closing control unit 38 narrows the distance between the arms 25 and the determination unit 37 determines that the pressure value on the rear side is not sufficient, the opening / closing control unit 38 slightly widens the distance between the arms 25 and after the gripping unit 21 is slightly lowered. The operation of narrowing the distance between the arms 25 is repeated until the determination unit 37 determines that the pressure value is within the predetermined range.
The spread interval of the arm 25 before the start of descent of the gripping unit 21 is appropriately set based on the three-dimensional information. However, even if there is an error, the shape of the part to be gripped of the article is generally known in advance from the three-dimensional information, so for example, the arm 25 is expanded by increasing the distance between the articles to be gripped by 5 mm. When the arm 25 is at a height of 10 mm above the planned gripping position, the arm 25 may be narrowed to search for the first contact.

Next, although partially overlapping, the flow of the picking method by the picking device 1 will be described. FIG. 8 is a flowchart showing the flow of the picking process.

A container C containing articles of the same shape stored in bulk is placed in the field of view of the picking device 1 (step S1).
The visual sensor 10 photographs the group of articles and outputs the three-dimensional information to the processing device 30 at predetermined time intervals (for example, 0.1 second intervals) (step S2).
The processing device 30 performs contour extraction from the input three-dimensional information and determines a collation figure whose size has been adjusted (step S3).
Further, the processing device 30 forcibly collates the articles from the article group using the collation figure, and calculates the center of the collation figure as the center position of the article (step S4).
The article to be lifted is determined, and the gripping unit 21 is lowered from above the center position of the article (step S5).

It is determined whether any of the tactile sensors 27 has detected contact with the article, that is, whether there has been a first contact (step S6), and if detected (step S6: Yes), the tactile sensor of the other arm 25. The gripping unit 21 is moved to the other arm 25 side until 27 detects a contact with the article (until a second contact is detected) (step S7).
The moving distance of the gripping unit 21 is calculated, and the center position is corrected based on the moving distance (step S8).
The gripping unit 21 is lowered from the vertical line at the corrected center position, the contact pressure on the front side of the arm 25 on one side is equal to or higher than the predetermined value A, and the contact pressure on the rear side is within a predetermined range as compared with the contact pressure on the front side. It is determined whether it is in B (step S9). If it is not within the predetermined range (step S9: No), the arm 25 spacing is widened by 2 mm, the gripping unit 21 is lowered by 2 mm, and the arm 25 spacing is narrowed again. That is, the search operation is performed (step S10). At this time, the interval is narrowed until at least the contact pressures on the two front sides reach a predetermined value A. Again, it is determined whether the difference between the contact pressure on the rear side and the contact pressure on the front side is within the predetermined range B (step S9). This is repeated until the two rear contact pressures are satisfied, the interval is slightly widened, the interval is slightly lowered, and the interval is narrowed until the front contact pressure is A or more. While searching, it is an operation of advancing the arm 25 to a certain deep position where stable grip is possible until a firm grip is achieved.

When it is within the predetermined range B (step S9: Yes), the gripping unit 21 is raised and placed on the conveyor (step S11), and the process returns to the start.

As described above, the picking device 1 and its picking method enable the introduction of a highly versatile and inexpensive system, and by performing on-site support, tuning and the like are not required, and prompt operation is realized. ..

The present invention is not limited to the above aspects. For example, if the article is extremely light or has a high elastic modulus, the article may be repelled by the first contact during normal operation. In such a case, a proximity detection function may be added to the contact sensor to reduce the contact speed. Alternatively, the gripping unit may be moved so as to search for the second contact by determining that the contact is close to each other (subsequent descent, search operation, etc. are normal, but the speed of various movements up to the pickup is slowed down. Is preferable).

According to the present invention, random picking of a general-purpose industrial robot by sight and touch can be realized at low cost, and introduction in small and medium-sized enterprises can be promoted.

1 Picking device 10 Visual sensor 20 Articulated robot 21 Gripping unit 22 Articulated unit 25 Arm 26 Base end drive unit 27 Tactile sensor 30 Processing device 31 3D information input unit 32 Sample graphic setting unit 33 Matching unit 34 Center calculation / correction unit 35 Articulated control unit 36 Tactile information input unit 37 Judgment unit 38 Open / close control unit

Claims (6)

A picking device that picks up articles of the same shape one by one from the sky from a container in which they are randomly stored.
A visual sensor that acquires three-dimensional information of a group of articles in a container,
Based on the 3D information of the article group acquired by the visual sensor, the articles are forcibly collated by one of a circle, a sphere, an ellipsoid, a square, a rectangle, a cube, a rectangular parallelepiped, and other sample figures, and each article is collated. A collation calculation means for calculating the center position and
A gripping unit consisting of at least two arms that have tactile sensors at the tips and grip the article by narrowing the distance.
A gripping control means that lowers the gripping unit from the sky at any center position and controls the movement of gripping and ascending the article by narrowing the arm spacing based on the contact detection of the article by the tactile sensor.
A picking device characterized by being equipped with. The tactile sensor is attached to the mounting surface at the tip of the arm as a flat pad body that grips the article by surface contact.
The picking device according to claim 1, wherein the normal direction of the mounting surface is 75 ° to 55 ° with respect to a line connecting the base end of the arm and the mounting surface. A movement control means for moving the gripping unit to the other arm side until the tactile sensor of the other arm detects the contact with the article after one of the tactile sensors detects the contact with the article.
A correction means that corrects the center position based on the movement distance of the gripping unit by the movement control means,
Equipped with
The picking device according to claim 1 or 2, wherein the picking control means controls an operation of gripping the article from a central position corrected by the correction means. The tactile sensor outputs the displacement value or the pressure value derived from the contact at at least two places, the front side and the rear side in the descending direction, respectively.
The output value from the tactile sensor is input, the output value on the front side is above a certain value, and the output value on the rear side asymptotically approaches the output value on the front side and falls within the predetermined range, or has not yet fallen. Equipped with a determination means for determining whether or not
The gripping control means narrows the arm spacing, and when it is determined by the determining means that it does not fit, the gripping control means repeats the operation of widening the arm spacing, lowering the gripping unit, and narrowing the arm spacing again until it is determined by the determining means. The picking device according to claim 1, 2 or 3, wherein the arm spacing is narrowed and motion control is performed to raise the gripping unit when it is determined by the determination means. Using a gripping unit in which arms arranged symmetrically with a tactile sensor at the tip grip and lift an article to be gripped evenly, and a visual sensor that acquires three-dimensional information of the article group. ,
This is a picking method in which articles of the same shape are picked up one by one from the sky from a container in which they are randomly stored.
A predetermined ellipsoidal shape, rectangular parallelepiped shape, and other sample figures are applied to the three-dimensional information of the article group acquired by the visual sensor, and the articles are forcibly collated to calculate the center position of each.
The article to be lifted is determined, the gripping unit is lowered from the sky above the center position of the article, and when one of the tactile sensors detects contact with the article, the tactile sensor of the other arm makes contact with the article. The feature is that the gripping unit is moved to the other arm side until it is detected, the center position is corrected based on the moving distance of the gripping unit, and the article is gripped and lifted from the vertical line of the corrected center position. Picking method to do. The tactile sensor outputs contact pressure at least at two locations, the front side and the rear side in the descending direction.
The operation of opening the arm, lowering the gripping unit, and closing the arm is repeated until the contact pressure on the front side is equal to or higher than the predetermined value and the contact pressure on the rear side gradually approaches the contact pressure on the front side and falls within the predetermined range. The picking method according to claim 5, wherein the article is then lifted.