JP2017217726A - robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique which improves accuracy in recognition of positions and attitudes of multiple work-pieces stored in bulk.SOLUTION: A robot 3 comprises: a work-piece recognition part 16 recognizing an inner space of a box 2 after dividing it into multiple cells 27, and then detecting a position of a work-piece 1 on the basis of positions of cells 27 in each of which a point cloud is detected using a distance image, among the divided cells 27; a distance measurement condition determining part 14 identifying cells 27 that are, among the divided cells 27, at least one of cells 27 excluding cells 27 labeled as "Occupied", as well as cells 27 each arranged at a position where the work-piece 1 has been detected by the work-piece recognition part 16, and then determining a distance measurement condition for a distance sensor 5 causing the number of cells 27 entering a sensing area of the distance sensor 5 to increase, among cells 27 labeled as "Occluded"; and distance measurement implementing part 15 moving the distance sensor 5 according to the distance measurement condition.SELECTED DRAWING: Figure 7

Description

  The present invention relates to a robot.

  As this type of technology, Patent Document 1 discloses a technology for measuring a relative position and posture between a component and the robot when the robot grips and assembles the component using an end effector such as a hand. doing. The robot is supposed to measure the position and orientation of a component using a two-dimensional image obtained by imaging the component with an imaging device and the distance data of the component measured by a distance sensor.

JP 2012-021958 A

  However, when a plurality of components are housed in a box in a stacked state, a blind spot for the imaging device and the distance sensor inevitably occurs. The blind spot may also change each time a part is removed from the box. Therefore, the technique of Patent Document 1 leaves room for improvement in the measurement of the position and orientation of parts.

  The objective of this invention is providing the technique which improves the recognition accuracy of the position and attitude | position of the some workpiece | work accommodated in the piled-up state.

  A robot that detects a desired object based on a sensing result using a distance measuring sensor and takes out the detected object with a hand. The robot divides a measurement area into a plurality of cells, recognizes the object, A detection process for executing the process of detecting the position of the object based on the position of the first cell, which is a cell where the object is detected using the sensing result of the distance measuring sensor, from among the divided cells And a second cell that is at least one of the divided cells, the cell excluding the first cell, and the cell at which the object is detected by the detection processing unit. The calculation means for calculating the movement position of the distance measuring sensor such that the number of the second cells in the sensing range of the distance measurement sensor among the second cells increases, and is calculated by the calculation means. In association with the movement position, the robot comprising a moving unit, the moving the distance measuring sensor is provided. According to the above method, the recognition accuracy of the positions and postures of the plurality of workpieces is improved in the third step.

  According to the present invention, the recognition accuracy of the positions and postures of the plurality of workpieces is improved.

It is a perspective view of the robot which takes out the workpiece | work accommodated in the box. It is a functional block diagram of a robot. It is a flowchart of operation | movement of a robot. It is a figure for demonstrating the flow which determines the first ranging condition. It is a figure which shows the shielding area which a ranging sensor cannot measure distance. It is a figure which shows the first ranging result by a ranging sensor. It is a figure which shows the ranging condition of the 2nd time by a ranging sensor. It is a figure which shows the point cloud obtained by the first ranging by the ranging sensor. It is a figure which shows the result of segmenting the first point cloud. It is a figure which shows the point cloud obtained by the ranging of the 2nd time by a ranging sensor. It is a figure which shows a mode that the first point cloud is integrated with the second point cloud. It is a figure which shows the result of having integrated the first point cloud into the second point cloud. It is a figure which shows the result of having segmented the 2nd point cloud after integration.

  Hereinafter, an embodiment will be described with reference to the drawings.

  FIG. 1 shows a robot 3 that takes out workpieces 1 one by one from a box 2 in which a plurality of workpieces 1 are accommodated in a bulk state.

  In the present embodiment, the plurality of workpieces 1 have the same shape and size.

  In the present embodiment, the box 2 has a bottomed rectangular tube shape that opens upward. However, the shape of the box 2 may be, for example, a bottomed cylindrical shape.

  The robot 3 includes a robot main body 4, a distance measuring sensor 5, a distance measuring sensor arm 6, a hand 7, and a hand arm 8.

  The distance measuring sensor 5 generates a distance image indicating the distance to the measurement target, and outputs the generated distance image to the robot body 4. The distance image is an image in which each pixel has depth information from the viewpoint position (position of the distance measuring sensor 5), and expresses distance data from the distance measuring sensor 5 to the measurement target. As the distance measuring sensor 5, an active distance measuring sensor that captures the reflected light of the laser light or slit light irradiated on the object with a camera and measures the distance by triangulation can be employed. As the distance measuring sensor 5, a time-of-flight distance measuring sensor that uses the flight time of light may be employed. Further, as the distance measuring sensor 5, a passive distance measuring sensor that calculates the depth of each pixel by triangulation from an image captured by a stereo camera may be employed.

  The distance measuring sensor arm 6 is an arm that holds the distance measuring sensor 5. The robot body 4 can freely control the position and posture of the distance measuring sensor 5 by driving the distance measuring sensor arm 6.

  The hand 7 is a part that holds or sucks the workpiece 1.

  The hand arm 8 is an arm that holds the hand 7. The robot body 4 can freely control the position and posture of the hand 7 by driving the hand arm 8.

  FIG. 2 is a functional block diagram of the robot 3. As shown in FIG. 2, the robot 3 further includes a sensor arm driving unit 9 and a hand arm driving unit 10.

  The robot body 4 includes a CPU 11 (Central Processing Unit) as a central processing unit, a read / write RAM 12 (Random Access Memory), and a read-only ROM 13 (Read Only Memory). Then, the CPU 11 reads and executes the control program stored in the ROM 13, so that the control program executes hardware such as the CPU 11, the distance measurement condition determination unit 14, the distance measurement execution unit 15, the workpiece recognition unit 16, the workpiece. It functions as the extraction execution unit 17.

  FIG. 3 shows a flowchart of the operation of the robot 3. As shown in FIG. 3, when the picking-up operation by the robot 3 is started, first, the distance measurement condition determination unit 14 measures the distance from the distance measurement sensor 5 to the plurality of workpieces 1 with the distance measurement sensor 5. Is determined (S100, first step). Here, the distance measurement condition means the position and orientation of the distance measurement sensor 5 when the distance measurement sensor 5 measures the distance from the distance measurement sensor 5 to the plurality of workpieces 1. Next, the distance measurement execution unit 15 measures the distances from the distance measurement sensor 5 to the plurality of workpieces 1 based on the distance measurement conditions determined in S100 (S110, second step). Next, the workpiece | work recognition part 16 recognizes the position and attitude | position of the some workpiece | work 1 based on the result of the measurement by the ranging sensor 5 (S120, 3rd step). And the workpiece removal execution part 17 takes out one workpiece | work 1 based on the result of recognition by the workpiece recognition part 16 (S130, 4th step). Then, the robot 3 determines whether all the workpieces 1 have been taken out (S140). If it is determined that all the workpieces 1 have been taken out (S140: YES), the processing is terminated. On the other hand, if it is determined that all the workpieces 1 have not been removed (S140: NO), the robot 3 returns the process to S100. That is, the robot 3 repeats the processing from S100 to S130 in order until all the workpieces 1 are taken out. Hereinafter, the operation of the robot 3 will be described in more detail.

(Operation of distance measurement condition determination unit 14)
Next, it will be described in detail that the distance measurement condition determination unit 14 determines the distance measurement condition in the first (N-1) th S100.

  As shown in FIG. 4, the distance measurement condition determination unit 14 calculates the three-dimensional coordinates of the geometric center point 19 of the inner bottom surface 18 of the box 2 in which a plurality of workpieces 1 are accommodated. Next, the distance measurement condition determination unit 14 generates a model of an arbitrary regular m-hedron 20, and the geometric center point 21 of the regular m-hedron 20 matches the geometric center point 19 of the inner bottom surface 18 of the box 2. Thus, the regular m-hedron 20 is positioned. Next, the distance measurement condition determination unit 14 calculates a straight line 23 that passes through the geometric center point 21 of the regular m-hedron 20 and is orthogonal to any one of the surfaces 22 constituting the regular m-hedron 20. Then, the distance measurement condition determining unit 14 defines a plurality of observation points 24 on the straight line 23 at predetermined intervals. The distance measurement condition determination unit 14 similarly defines a plurality of observation points 24 for the other surfaces 22 constituting the regular m-plane 20. The distance measuring sensor 5 is arranged at one of the observation points 24 and measures the distance from the observation point 24 toward the geometric center point 21 of the regular m-hedron 20. Accordingly, the distance measurement condition determination unit 14 sets the distance measurement sensor 5 to any observation point 24 as the distance measurement condition of the distance measurement sensor 5, and the posture of the distance measurement sensor 5 changes from the observation point 24 to the regular m-hedron. The RAM 12 stores a plurality of distance measuring conditions indicating that the posture is toward the 20 geometric center points 21. Then, the distance measurement condition determination unit 14 narrows down the plurality of distance measurement conditions stored in the RAM 12 according to the following conditions. That is, the distance measurement condition determination unit 14 deletes a distance measurement condition from the RAM 12 where the position of the distance measurement sensor 5 is not within the work area of the robot 3. In addition, the distance measurement condition determination unit 14 deletes the distance measurement condition having no inverse kinematics solution from the RAM 12. Further, the distance measurement condition determination unit 14 may cause the robot 3 to physically interfere with the external environment when the distance measurement sensor 5 is moved so as to satisfy the distance measurement condition, or so-called self-interference may occur. Then, the distance measurement condition is deleted from the RAM 12. Further, the distance measurement condition determination unit 14 deletes the distance measurement condition from the RAM 12 when the movement of the distance measurement sensor 5 to satisfy the distance measurement condition is out of the movable range of the robot 3. Next, the distance measurement condition determination unit 14 selects a distance measurement condition that allows easy measurement of the inner bottom surface 18 of the box 2 having a known shape, among the distance measurement conditions stored in the RAM 12. FIG. 5 shows a state in which the distance measuring sensor 5 measures the inner bottom surface 18 of the box 2. As shown in FIG. 5, it is not always possible to measure the inner bottom surface 18 of the box 2 with the distance measuring sensor 5 from above the box 2. Therefore, when the distance measurement sensor 5 measures the inner bottom surface 18 of the box 2, there is a shielding area une that cannot be measured due to the presence of the side wall 25 of the box 2. Further, when the obstacle 26 exists above the box 2, when the distance measuring sensor 5 measures the inner bottom surface 18 of the box 2, a shielding area une that cannot be measured is generated behind the obstacle 26. Therefore, the distance measurement condition determination unit 14 selects a distance measurement condition that minimizes the area of the shielding region une. When there are a plurality of ranging conditions that minimize the area of the shielding area une, the ranging condition determining unit 14 sets the inner bottom surface 18 of the box 2 as the focus of the ranging sensor 5 as the first priority condition. The one closer to the distance is selected, and as a second priority condition, the normal passing through the geometric center point 19 of the inner bottom surface 18 of the box 2 and orthogonal to the inner bottom surface 18 and the geometry of the inner bottom surface 18 of the box 2 are selected. The smaller angle formed by the target center point 19 and the straight line connecting the distance measuring sensor 5 is selected. The distance measurement condition determination unit 14 records the distance measurement condition determined by the selection in the RAM 12.

(Operation of the distance measurement execution unit 15)
Next, in the first (N-1) th time S110, the distance measurement execution unit 15 determines from the distance measurement sensor 5 to the plurality of workpieces 1 based on the distance measurement condition determined by the distance measurement condition determination unit 14 in S100. Measuring the distance with the distance measuring sensor 5 will be described in detail.

  First, the distance measurement execution unit 15 controls the sensor arm drive unit 9 to change the position and orientation of the distance measurement sensor 5 so as to satisfy the distance measurement condition determined by the distance measurement condition determination unit 14. Next, the distance measurement execution unit 15 transmits a distance measurement command to the distance measurement sensor 5. When receiving the distance measurement command, the distance measurement sensor 5 performs distance measurement to generate a distance image, and transmits the generated distance image to the robot body 4. The distance measurement execution unit 15 stores the received distance image in the RAM 12.

(Operation of workpiece recognition unit 16)
Next, it will be described in detail that the workpiece recognition unit 16 recognizes the positions and postures of the plurality of workpieces 1 based on the distance image in the first (N-1th) S120.

  First, the workpiece recognition unit 16 reads a distance image stored in the RAM 12 and converts the distance image into a set of points having three-dimensional coordinates. A set of points having three-dimensional coordinates is generally referred to as a point cloud. The workpiece recognition unit 16 stores the first point cloud in the RAM 12. The workpiece recognition unit 16 segments the initial point cloud and compares a plurality of points belonging to each segment with an AABB (Axis Aligned Bounding Box) of the workpiece 1. And the workpiece | work recognition part 16 recognizes the position and attitude | position of each workpiece | work 1 based on the some point which belongs to the segment with high coincidence with AABB of the workpiece | work 1 as a result of the said comparison. At this time, the workpiece recognition unit 16 preferably recognizes the positions and postures of all the workpieces 1 accommodated in the box 2. However, instead of recognizing the positions and postures of all the workpieces 1 accommodated in the box 2, the workpiece recognition unit 16 replaces at least one position among the plurality of workpieces 1 accommodated in the box 2. And recognizing posture.

(Operation of work removal execution unit 17)
Next, in the first (N-1th) S130, the workpiece removal execution unit 17 extracts one workpiece 1 based on the recognition result by the workpiece recognition unit 16, as described above.

(Operation of distance measurement condition determination unit 14)
Next, it will be described in detail that the distance measurement condition determination unit 14 determines the distance measurement condition in the second (Nth) S100.

  First, as shown in FIG. 6, the distance measurement condition determining unit 14 divides the internal space of the box 2 into a lattice shape to define a plurality of cells 27 arranged in a three-dimensional shape. Then, the distance measurement condition determining unit 14 reads the initial point cloud stored in the RAM 12 and stores the initial point cloud in the plurality of cells 27. Then, the label “Occupied” is attached to the cell 27 in which at least one point in the initial point cloud exists. The cell 27 (first cell 30) labeled “Occupied” is identified by the alphabet “o” in FIG. Then, the distance measurement condition determination unit 14 determines that the cell 27 existing in an area where the distance measurement sensor 5 cannot measure the distance due to the presence of the cell 27 labeled “Occupied” as viewed from the distance measurement sensor 5. Label it “Occluded”. An area that cannot be measured by the distance measuring sensor 5 due to the presence of the cell 27 labeled “Occupied” is hereinafter also referred to as a blind spot. The cell 27 labeled “Occluded” is identified by the alphabet “x” in FIG. Note that the distance measuring sensor 5 shown in FIG. 6 has a position and orientation that satisfy the first distance measuring condition. According to FIG. 6, it can be seen that the blind spot 28 was generated under the first ranging condition.

  In addition, the distance measurement condition determination unit 14 rewrites the label of the cell 27 corresponding to the workpiece 1 taken out by the workpiece removal execution unit 17 for the first time to “Occluded”. In the present embodiment, since the work 1 on the left side of the page is taken out for the first time, the label of the cell 27 on the left side of the page is rewritten from “Occupied” to “Occluded” as shown by reference numeral 29 as shown in FIG. It has been.

  Then, the distance measurement condition determination unit 14 determines the distance measurement condition of the distance measurement sensor 5 so that more distances can be measured in the cells 27 (second cells 31) labeled “Occluded”. Specifically, in the Nth time, the distance measurement condition determination unit 14 has an area (cell 27) that has become the blind spot 28 for the first time and an area (cell 27) indicated by reference numeral 29 in which the work 1 taken out for the first time exists. The distance measuring sensor 5 is preferably maximized so that the area (cell 27) in which the distance measuring sensor 5 can measure the distance increases in comparison with the previous time. Determine the distance measurement conditions. FIG. 7 shows the distance measuring sensor 5 having a position and orientation that satisfy the distance measuring conditions determined in this way. According to this, since it becomes possible to grasp the distance measurement object evenly, the accuracy with which the work recognition unit 16 recognizes the positions and postures of the plurality of works 1 for the Nth time is improved. In particular, by treating the area where the work 1 taken out at the first time existed in the same manner as the blind spot 28, the work recognition unit 16 can fully grasp the change in the state in the box 2 as the work 1 is taken out. Will be recognized.

(Operation of the distance measurement execution unit 15)
Next, in the second (Nth) S110, the distance measurement execution unit 15 performs a plurality of works 1 from the distance measurement sensor 5 based on the distance measurement conditions determined by the distance measurement condition determination unit 14 in the second S100. Is measured by the distance measuring sensor 5. The distance measurement execution unit 15 stores the second distance image in the RAM 12.

(Operation of workpiece recognition unit 16)
Next, it will be described in detail that the workpiece recognition unit 16 recognizes the positions and postures of the plurality of workpieces 1 based on the distance image in the second (Nth) S120.

  First, the workpiece recognition unit 16 reads the second distance image stored in the RAM 12 and converts the second distance image into a set of points having three-dimensional coordinates. The work recognition unit 16 stores the second point cloud in the RAM 12. In this state, the first point cloud and the second point cloud are stored in the RAM 12.

  In the second S120, the work recognition unit 16 first integrates the first point cloud into the second point cloud (integration step). Second, the workpiece recognition unit 16 performs segmentation on the second point cloud after integration (segmentation process). Third, the position and orientation of the workpiece 1 are recognized based on the segmentation result (recognition step). Hereinafter, the integration process, the segmentation process, and the recognition process will be described in detail in order.

(Integration process)
First, an integration process for integrating the first point cloud into the second point cloud will be described. If the first point cloud is integrated into the second point cloud, a point cloud with few blind spots obtained by ranging from different ranging conditions can be obtained, so that the workpiece recognition unit 16 can improve the recognition accuracy of the workpiece 1 Contribute. However, it is not reasonable to integrate all the first point clouds into the second point cloud. This is because the first point cloud is not always consistent with the plurality of workpieces 1 at the second time by taking out the first workpiece 1. Therefore, in this embodiment, the first point cloud can be divided into those that can be integrated into the second point cloud and those that cannot be integrated into the second point cloud. It will be integrated into the point cloud. This segmentation is performed for each segment obtained by segmenting the initial point cloud.

FIG. 8 shows the first point P _N-1 (j). FIG. 9 shows a plurality of segments S _N-1 (t) obtained as a result of the work recognition unit 16 performing segmentation on the first point P _N-1 (j). FIG. 10 shows the second point P _N (i). As shown in FIG. 9, a plurality of points P _N-1 (j) belong to one segment S _N-1 (t). In other words, one segment S _N-1 (t) is composed of a plurality of points P _N-1 (j).

The workpiece recognizing unit 16 determines whether each of the initial points P _N-1 (j) can be integrated with the second point P _N (i) for each initial segment S _N-1 (t). judge. Therefore, as the counter used for the determination, the counters Near _N-1 (t) and Far _N-1 (t) are defined for each of the initial segments S _N-1 (t), and the initial values are defined. Zero. In FIG. 11, in order to facilitate understanding of the explanation, all of the first segment S _N-1 (t) and the second point P _N (i) are displayed in an overlapping manner. In FIG. 11, the second point P _N (i) is indicated by a bold circle.

First, the workpiece recognition unit 16 includes a second time at the first point P _N (i) Point P _N (1) is, by comparing the, all the initial point P _N-1 (j), the first point P _N-1 of the (j), to identify the nearest point P _N-1 (j) to a point P _N (1). The identified point P _N-1 (j) is set as the point P _N-1 (k). Then, the workpiece recognition unit 16 calculates the distance d between the point P _N (1) and the point P _N-1 (k), and the segment to which the point P _N-1 (k) belongs according to the distance d. Increment the counter of S _N-1 (t). Specifically, when the distance d is equal to or smaller than a predetermined value, the workpiece recognition unit 16 determines Near _N-1 (t) of the segment S _N-1 (t) to which the point P _N-1 (k) belongs. Increment by one. On the other hand, when the distance d exceeds the predetermined value, the workpiece recognition unit 16 increments Far _N-1 (t) of the segment S _N-1 (t) to which the point P _N-1 (k) belongs by one. To do. Next, the workpiece recognition unit 16 performs the same process for the second and subsequent points P _N (i) for the second time.

Next, the workpiece recognizing unit 16 determines that the first point P _N-1 (j) belonging to the segment S _N-1 (1) which is the first first segment S _N-1 (t) is the second point P. _Judge whether it can be integrated into (i). Specifically, the workpiece recognition unit 16 performs the above determination based on Near _N-1 (1) and Far _N-1 (1) which are variables of the segment S _N-1 (1). That is, the workpiece recognition unit 16, Near when _N-1 (1) and Far _N-1 the sum of (1) is less than a predetermined value, the point P _N-1 belonging to the segment S _N-1 (1) ( j) is equivalent to the blind spot in the second ranging, and the point P _N-1 (j) belonging to the segment S _N-1 (1) is integrated into the second point P _N (i) as it is (added) , Add, transfer). On the other hand, when the total value of Near _N-1 (1) and Far _N-1 (1) exceeds a predetermined value, the workpiece recognition unit 16 further branches the condition, and Far _N-1 (1) / Near _{N -1} (1) is equal to or less than a predetermined value, it is assumed that the workpiece 1 corresponding to the point P _N-1 (j) belonging to the segment S _N-1 (1) does not move when the workpiece 1 is taken out for the first time. The point P _N-1 (j) belonging to S _N-1 (1) is integrated into the second point P _N (i) as it is, while Far _N-1 (1) / Near _N-1 (1 ) Exceeds a predetermined value, or when Near _N-1 (1) is zero, work 1 corresponding to point P _N-1 (j) belonging to segment S _N-1 (1) is the first work The point P _N-1 (j) belonging to the segment S _N-1 (1) is not integrated into the second point P _N (i) as being taken out at the time of taking out one. Next, the workpiece recognizing unit 16 performs the same process for the first and subsequent segments S _N-1 (t). FIG. 12 shows the result of integration. In FIG. 12, a plurality of points P _N (i) overlap at the same coordinates. This duplication itself is preferably left as it is in terms of simplification of processing, and in terms of speeding up of processing, it is preferable to eliminate the duplication state and leave only one of them. In the present embodiment, priority is given to simplification of processing, and the overlapping state is left as it is.

(Segmentation process)
Next, the workpiece recognition unit 16 performs segmentation on the second point P _N (i) after integration shown in FIG. The second segment S _N (t) obtained by this segmentation is shown in FIG.

(Recognition process)
Next, the workpiece recognition unit 16 recognizes the positions and postures of the plurality of workpieces 1 based on the second segment S _N (t). In general, the calculation cost required for recognizing the position and orientation of the workpiece 1 is high, and it is from the aspect of calculation cost that the position and orientation of all the workpieces 1 are recognized every time based on the segment S _N (t). It is not preferable. Therefore, in this embodiment, instead of recognizing the positions and postures of all the workpieces 1 based on the segment S _N (t) every time, the portion of the previous recognition result that can be diverted is diverted. . Hereinafter, it will be described in detail which of the previous recognition results is used to determine which work 1 is recognized.

First, the workpiece recognition unit 16 determines, for each segment S _N (t), whether the recognition result of the workpiece _N-1 (r) obtained for the first time can be used for the second recognition process. Therefore, as counters used for judgment, the counters Near _N-1 (r) and Far _N-1 (r) are defined for each of the workpieces _N-1 (r) whose position and orientation are recognized for the first time. The initial value is set to zero.

Next, workpiece recognition unit 16 calculates the AABB of the first segment S _N (t) is a segment S _N (1) in the second time, the AABB segment S _N (1), the first workpiece _{N- 1} Determine whether there is a collision with all AABBs in (r). Then, the workpiece recognition unit 16 identifies the first workpiece _N-1 (r) that collides with the AABB of the segment S _N (1), and identifies the identified workpiece _N-1 (r) as the workpiece _N-1 (s). To do. The workpiece recognition unit 16, segment S and _N points belonging to (1) P _N (i) Point P _N (1) is the first point P _N (i) of the work _N-1 (s) And the counter of the workpiece _N-1 (s) is incremented according to the shortest distance d. Specifically, the work recognizing unit 16 increments Near _N-1 (s) of the work _N-1 (s) when the shortest distance d is equal to or less than a predetermined value. On the other hand, when the shortest distance d exceeds the predetermined value, the workpiece recognition unit 16 increments Far _N-1 (s) of the workpiece _N-1 (s). When there are a plurality of initial works _N-1 (r) colliding with AABB in the segment S _N (1), the other works _N-1 (r) are processed in the same manner.

Next, the workpiece recognition unit 16 determines whether or not the recognition result of the workpiece _N-1 (s) can be used for the second recognition. Specifically, the above determination is performed based on Near _N-1 (s) and Far _N-1 (s), which are variables of the work _N-1 (s). That is, the workpiece recognition unit 16 can divert the recognition result of the workpiece _N-1 (s) for the second recognition when Far _N-1 (s) / Near _N-1 (s) is equal to or less than a predetermined value. Therefore, the recognition of the position and posture of the workpiece 1 based on the point point P _N (i) included in the first segment S _N (t) at the second time is omitted. On the other hand, when Far _N-1 (s) / Near _N-1 (s) exceeds a predetermined value or Near _N-1 (s) is zero, the workpiece recognizing unit 16 sets the workpiece _N-1 ( The recognition result of s) is determined to be unusable for the second recognition, and the position and posture of the workpiece 1 based on the point point P _N (i) included in the first segment S _N (t) at the second time Recognize Next, the workpiece recognition unit 16 performs the same process for the second and subsequent segments S _N (t) for the second time.

(Operation of work removal execution unit 17)
Next, in the second (N-th) S130, the workpiece take-out execution unit 17 takes out one workpiece 1 based on the recognition result by the workpiece recognition unit 16.

  Although the preferred embodiment has been described above, the above embodiment has the following features.

  The robot 3 detects the desired workpiece 1 (target object) based on the distance image (sensing result using the distance measuring sensor 5) and takes out the detected workpiece 1 with the hand 7. ) Is divided into a plurality of cells 27, and a cell at a position where a point cloud (object) is detected from the divided cells 27 using a distance image (sensing result of the distance measuring sensor 5). 27, the workpiece recognition unit 16 (detection) that executes processing for detecting the position of the workpiece 1 based on the position of the first cell 30 (the first cell 30 is the cell 27 labeled “Occupied”). A second cell that is at least one of the processing unit), the cell 27 excluding the first cell 30 among the divided cells 27, and the cell 27 at the position where the workpiece 1 is detected by the workpiece recognition unit 16. ( The cell 27) labeled “Occluded” is identified, and the number of the second cells 31 in the sensing range of the distance measuring sensor 5 among the second cells 31 is increased as compared with the previous distance measurement. A distance measurement condition determination unit 14 (calculation means) that calculates the distance measurement condition (movement position) of the distance measurement sensor 5 and the distance measurement sensor 5 are moved in association with the movement position calculated by the distance measurement condition determination unit 14. A distance measuring execution unit 15 (moving means). According to this, since it becomes possible to grasp the distance measurement object evenly, the accuracy with which the work recognition unit 16 recognizes the positions and postures of the plurality of works 1 for the Nth time is improved. In particular, by treating the area where the work 1 taken out at the first time existed in the same manner as the blind spot 28, the work recognition unit 16 can fully grasp the change in the state in the box 2 as the work 1 is taken out. Will be recognized.

  A control method for a robot that takes out the workpieces one by one from a box in which a plurality of workpieces are stored in a stacked state, and when measuring distances from the ranging sensors to the plurality of workpieces by a ranging sensor. Based on the first step of determining the position and orientation of the distance measuring sensor and the result of the determination in the first step, the distance from the distance measuring sensor to the plurality of workpieces is measured by the distance measuring sensor. Based on the second step, the third step for recognizing the positions and postures of the plurality of workpieces based on the measurement result in the second step, and the recognition result in the third step, And a fourth step of sequentially taking out the workpiece, and in the first step of the Nth time, a blind spot is formed in the second step of the (N-1) th time. The area where the distance measurement sensor can measure the distance is the previous area of the specific area including the area where the workpiece has been taken out in the fourth step of the (N-1) th time. There is provided a control method for determining the position and orientation of the distance measuring sensor so as to increase as compared with the distance.

DESCRIPTION OF SYMBOLS 1 Work 2 Box 3 Robot 4 Robot main body 5 Distance sensor 6 Distance sensor arm 7 Hand 8 Hand arm 9 Sensor arm drive part 10 Hand arm drive part 14 Distance measurement condition determination part 15 Distance execution part 16 Work recognition part 17 Work Extraction execution unit 27 cell 28 blind spot 30 first cell 31 second cell

Claims (1)

A robot that detects a desired object based on a sensing result using a distance sensor and takes out the detected object with a hand,
After dividing the measurement area into a plurality of cells and recognizing, the position of the first cell, which is the cell where the object is detected using the sensing result of the distance measuring sensor, is selected from the divided cells. Detection processing means for executing processing for detecting the position of the object based on the above;
Among the divided cells, a second cell that is at least one of a cell excluding the first cell and a cell at a position where the object is detected by the detection processing unit is specified, and the first cell A calculating means for calculating a movement position of the distance measuring sensor such that the number of the second cells entering the sensing range of the distance measuring sensor among the two cells increases;
Moving means for moving the distance measuring sensor in association with the moving position calculated by the calculating means;
Robot equipped with.

Images