Gripping device for a robot for gripping an object, method for gripping an object using a gripping device of a robot

Abstract

The invention relates to a gripping device (1) for a robot (3) for gripping an object (5), the device having a main body (7), a plurality of fingers (9) which are movable relative to the main body (7), and three or more cameras (11) for detecting the object, the cameras (11) each having different detection ranges, and it being possible for a two-dimensional image of the object (5) to be generated by each of the cameras (11), characterised by a control unit which is configured to generate a three-dimensional representation of the object (5) on the basis of the two-dimensional images generated by the cameras (11) and to move the fingers (9) on the basis of the three-dimensional representation of the object (5) so that the object (5) is gripped.

Description

[0001] Gripping device for a robot for gripping an object, method for gripping an object with a gripping device of a robot

[0002] The invention relates to a gripping device for a robot for gripping an object. Furthermore, the invention relates to a method for gripping an object with a gripping device of a robot.

[0003] With the increasing automation of industrial operations, particularly in production and logistics, as well as in healthcare and other sectors, a growing number of robots are being deployed to perform, facilitate, and optimize work. Such robots are used for a variety of purposes and can incorporate various interchangeable tools on the so-called end effector to fulfill their purpose. In addition to welding fixtures, various gripping devices can be used as tools to grasp objects and, in particular, to change their spatial position and / or orientation. A camera can be used to precisely localize the object and ultimately optimize gripping.

[0004] A well-known challenge in robot-assisted gripping of an object in the state of the art is the so-called occlusion problem. Occlusion refers to the partial or complete blocking of a camera's view, particularly by components of the robot itself, in particular by the gripping device or a robot arm segment. Thus, occlusion results in more difficult capture and ultimately suboptimal gripping of the object. As the gripping device or camera approaches the object closer, the occlusion can become more severe. Some attempts are made to solve the occlusion problem by increasing the number of cameras, i.e., by using a camera array, in order to achieve more comprehensive capture of the object. However, any individual camera in a camera array can also be subject to occlusion. In addition, multiple cameras generally pose another challenge: increased costs.Due to economic limitations, the occlusion problem cannot be solved simply by increasing the number of cameras.

[0005] Against this background, the task arises of providing a cost-effective solution that enables improved detection and thus optimized gripping of an object by means of a gripping device. The task is solved by a gripping device for a robot for gripping an object, comprising a base body, a plurality of fingers movable relative to the base body, and three or more cameras for detecting the object, wherein the cameras each have different detection ranges and wherein a two-dimensional image of the object can be generated by means of the cameras, and comprising a control unit that is configured to generate a three-dimensional representation of the object based on the two-dimensional images generated by the cameras and to move the fingers based on the three-dimensional representation of the object such that the object is gripped.

[0006] The gripping device has a base body and several fingers that can be moved relative to the base body. The fingers can be directly connected to the base body for rotation and have several articulated finger segments. This has the particular advantage of imitating a human hand. The gripping device can therefore be used in particular in humanoid robots. Furthermore, the gripping device comprises three or more cameras for capturing the object. The cameras each have a different detection range. This results in the advantage that with an increasing number of cameras, the overall detection range increases and, in particular, the object can be captured more comprehensively. The cameras can each generate a two-dimensional image of the object, in particular simultaneously.According to the invention, the gripping device has a control unit configured to generate a three-dimensional representation of the object based on the two-dimensional images generated by the cameras. According to the invention, this provides the advantage that even areas of the object that lie outside the respective detection range of the cameras can be represented three-dimensionally, in particular in color. Based on the three-dimensional representation of the object, the fingers can be moved in such a way that the object is grasped.

[0007] A three-dimensional representation of the object within the meaning of the invention is understood to mean, in particular, a data set that describes, in particular, the surface and orientation of the object. The three-dimensional representation can, for example, be a point cloud, a mesh, or a depth map, on the basis of which gripping can be optimized and carried out more reliably.

[0008] According to an advantageous embodiment of the invention, it is provided that the

[0009] The control unit is configured to generate the three-dimensional representation of the object using a generative artificial intelligence method, in particular a Neural Radiance Field method or a NeuralMVS method. Using a generative artificial intelligence method, the advantage can be achieved that the three-dimensional representation of the object also contains aspects that relate to areas of the object not captured by the cameras. For example, the generative artificial intelligence method can be used to generate a configuration of a rear side or base of the object that is not visible to the cameras.

[0010] In a preferred embodiment of the invention, at least one, preferably several, of the cameras are arranged so as to be movable relative to one or more other cameras. By means of an absolute or relative mobility of at least one camera, the advantage of variable detection areas of the gripping device can arise, whereby the object can be detected in an optimized manner. The two-dimensional images can be generated in particular during a movement of the gripping device - i.e. also during a relative movement of at least one camera with respect to at least one other camera. If the three-dimensional representation of the object is generated using a generative artificial intelligence method, this has the advantage that knowledge of the orientation and / or arrangement of the multiple cameras is not required.Therefore, there is no need to record the camera orientation and / or arrangement, resulting in reduced equipment costs.

[0011] An advantageous embodiment of the invention provides that at least one, preferably several, of the cameras is arranged on a finger. By attaching at least one camera to at least one movable finger, at least two cameras can be moved relative to each other. Arranging a camera on a finger also offers the advantage that detection can take place in the immediate vicinity of the surfaces, particularly the finger, that come into contact with the object when grasping it. Cameras arranged in this way are less likely to be exposed to the occlusion problem because they are more likely to have a direct view of the object.

[0012] According to a preferred embodiment of the invention, it is provided that one, preferably several, of the cameras are arranged in a cavity in the finger. The fingers can have several cavities. Each cavity can comprise one or more cameras. By arranging one or more cameras in a finger, the gripping device can be made more compact. Preferably, one or more cameras are arranged in each of several fingers. In an advantageous embodiment of the invention, it is provided that at least one, preferably several, of the cameras are arranged on the base body. The one or more cameras are preferably arranged on a side of the base body facing the fingers and the object. Preferably, at least one camera is movably connected to the base body by means of a joint, i.e., in an articulated manner. The base body can be flat or trough-shaped or have a flat or trough-shaped section.The cameras can also be arranged - at least partially - in an edge area of ​​the base body, which has the advantage of as many different detection areas and viewing angles as possible.

[0013] A preferred embodiment of the invention provides that one, preferably several, of the cameras is arranged in a cavity of the base body. By arranging one or more cameras in a cavity, the gripping device can be designed more compactly, since it is not necessary for the cameras to protrude from a surface of the base body. Arranging the one or more cameras in a cavity also reduces the likelihood of occlusion problems caused by the cameras. The base body can have several cavities, with one or more cameras being arranged in each cavity.

[0014] According to a preferred embodiment of the invention, the base body has a surface facing the object with a central section, wherein one of the cameras is arranged on a central line arranged perpendicular to the central section, wherein the detection area of ​​the camera is formed symmetrically around the central line. The central line is preferably arranged parallel to a rotation axis of the gripping device.

[0015] In a further embodiment of the invention, several other cameras are arranged symmetrically, in particular concentrically, around the central line. The other cameras, in particular three or five, can be arranged in particular along a circle, particularly preferably along a circumferential direction of the circle. The other cameras can be evenly spaced along the circumferential direction. Furthermore, the other cameras can be arranged along a quadrilateral.

[0016] A preferred embodiment of the invention provides that the fingers each have a gripping surface, wherein the gripping surfaces of at least two fingers are arranged in parallel. In other words, the object can be gripped at two parallel surface sections. This results in the technical effect that a large clamping force can be built up easily. Furthermore, there is the advantage that gripping can be more secure because the risk of slipping when gripping is lower due to the higher clamping force. Gripping surfaces can be two- or three-dimensional, and at least partially curved and / or faceted. The gripping surfaces can have an unevenness or roughness. Furthermore, the gripping surfaces can each comprise a separately formed gripping layer for more secure gripping, for example made of silicone.

[0017] A further subject matter of the invention is a method for gripping an object with a gripping device of a robot, wherein the gripping device has a base body, a plurality of fingers movable relative to the base body and three or more cameras for detecting the object, wherein the cameras each have different detection areas and wherein a two-dimensional image of the object is generated by means of the cameras, wherein a three-dimensional representation of the object is generated on the basis of the two-dimensional images generated by the cameras and the fingers are moved on the basis of the three-dimensional representation of the object in such a way that the object is gripped.

[0018] The method according to the invention can achieve the same technical effects and advantages that have already been described in connection with the gripping device according to the invention.

[0019] The preferred embodiments and features explained in connection with the gripping device can also be used, alone or in combination, in the method according to the invention.

[0020] Further details and advantages of the invention will be explained below with reference to the exemplary embodiments illustrated in the drawings. Herein:

[0021] Fig. 1 schematically shows an embodiment of a gripping device according to the invention for a robot for gripping an object;

[0022] Fig. 2 schematically shows another embodiment of the inventive

[0023] Gripping device for a robot for gripping another object; Fig. 3 schematically shows three possible camera arrangements; and

[0024] Fig. 4 schematically shows a detailed view of a third embodiment of the gripping device, wherein a camera is arranged in a cavity of a base body.

[0025] Fig. 1 schematically shows a first embodiment of a gripping device 1 according to the invention for a robot 3 (not shown in Fig. 1) for gripping a first object 5. The gripping device 1 comprises a base body 7, three fingers 9 movable relative to the base body 7, and seven cameras 11 for detecting the object 5, wherein the cameras 11 each have different detection areas and wherein a two-dimensional image of the object 5 can be captured by means of the cameras 11. According to the invention, the gripping device 1 has a control unit 13 which is configured to generate a three-dimensional representation of the object 5 on the basis of the two-dimensional images captured by the cameras 11 and to move the fingers 9 on the basis of the three-dimensional representation of the object 5 such that the object 5 is gripped.

[0026] Fig. 2 schematically shows a further embodiment of the gripping device 1 according to the invention for a robot 3 for gripping a further object 5'. The further embodiment according to Fig. 2 has a base body 7 and two fingers 9 movable relative to the base body 7. The two fingers 9 each have four finger segments 9', wherein the most distally arranged finger segment 9' of the two fingers 9 is designed as a contact and clamping element for gripping the further object 5'.

[0027] Fig. 3 shows three possible camera arrangements. All three camera arrangements represent alternative designs to the camera arrangement shown in Fig. 1, in which the three cameras 11 are arranged in a row.

[0028] Fig. 3a shows a triangular camera arrangement. Fig. 3b also shows a quadrangular camera arrangement. In addition, Fig. 3c shows a star-shaped camera arrangement, wherein one of the cameras 11 is arranged as the central camera 11z and three further cameras 11 as planetary cameras 11p. The planetary cameras 11p are evenly spaced with respect to a circumferential direction of a circle formed around the central camera 11z. Fig. 4 shows a schematic detailed view of a third embodiment of the gripping device 1, wherein a camera 11 is arranged in a cavity K of the base body. The base body 7 has a surface facing a third object 5" with a central section A. The camera 11 is arranged on a central line Z arranged perpendicular to the central section A, wherein the detection area E of the camera 11 is formed symmetrically around the central line Z.

[0029] List of reference symbols

[0030] 1 gripping device

[0031] 3 robots

[0032] 5 Object

[0033] 5' Additional Object

[0034] 5" Third Object

[0035] 7 basic bodies

[0036] 9 fingers

[0037] 9' finger segments

[0038] 11 Camera

[0039] 11p planetary camera

[0040] 11z central camera

[0041] 13 Control unit

[0042] E Detection range

[0043] K Cavity

[0044] Z Central Line

Claims

Patent claims 1. Gripping device (1) for a robot (3) for gripping an object (5), with a base body (7), a plurality of fingers (9) which can be moved relative to the base body (7), and three or more cameras (11) for detecting the object, wherein the cameras (11) each have different detection areas and wherein a two-dimensional image of the object (5) can be generated by means of the cameras (11), characterized by a control unit (13) which is configured to generate a three-dimensional representation of the object (5) on the basis of the two-dimensional images generated by the cameras (11) and to move the fingers (9) on the basis of the three-dimensional representation of the object (5) in such a way that the object (5) is gripped.

2. Gripping device (1) according to claim 1, characterized in that the control unit is configured to generate the three-dimensional representation of the object (5) using a generative artificial intelligence method, in particular a Neural Radiance Field method or a NeuralMVS method.

3. Gripping device (1) according to one of the preceding claims, characterized in that at least one, preferably several, of the cameras (11) are arranged to be movable relative to one or more other cameras.

4. Gripping device (1) according to one of the preceding claims, characterized in that at least one, preferably several, of the cameras (11) are arranged on a finger.

5. Gripping device according to claim 4, characterized in that one, preferably several, of the cameras (11) are arranged in a cavity in the finger.

6. Gripping device (1) according to one of the preceding claims, characterized in that at least one, preferably several, of the cameras (11) are arranged on the base body (7).

7. Gripping device (1) according to claim 6, characterized in that one, preferably several, of the cameras (11) are arranged in a cavity of the base body.

8. Gripping device (1) according to one of the preceding claims, characterized in that the base body (7) has a surface facing the object (5) with a central section, wherein one of the cameras (11) is arranged on a central line (Z) arranged perpendicular to the central section, wherein the detection range of the camera (11) is formed symmetrically around the central line (Z).

9. Gripping device (1) according to claim 8, characterized in that several other cameras (11) are arranged symmetrically, in particular concentrically, around the central line (Z).

10. A method for gripping an object (5) with a gripping device (1) of a robot (3), wherein the gripping device (1) has a base body (7), a plurality of fingers (9) movable relative to the base body (7), and three or more cameras (11) for detecting the object (5), wherein the cameras (11) each have different detection areas and wherein a two-dimensional image of the object (5) is generated by means of the cameras (11), wherein a three-dimensional representation of the object (5) is generated on the basis of the two-dimensional images generated by the cameras (11), and based on the three-dimensional representation of the object (5), the fingers are moved in such a way that the object is gripped.

Images