Parallel-jawed robotic picking method, apparatus, device and product

Abstract

The application provides a parallel gripper robot picking method, device, equipment and product, and relates to the technical field of automation, and comprises the following steps: acquiring a point cloud image of a target scene, obtaining a first picking pose based on the point cloud image and a pre-trained picking pose prediction model; determining a second picking pose for a target object and a score from the first picking pose; adjusting each second picking pose in turn according to the scores of each second picking pose from high to low until determining that there is no overlapping part between the gripper of the parallel gripper robot and the point cloud image in the third picking pose after the adjustment of a certain second picking pose; and controlling the parallel gripper robot to pick the target object from the target scene according to the third picking pose. Therefore, the scene features of the target cluttered scene can be accurately expressed through the point cloud image in the visual feedback method, and the parallel gripper robot can adapt to complex cluttered environments and accurately pick through the adjustment of the candidate picking pose, thereby improving the picking accuracy and picking efficiency of the robot in the cluttered environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation, and in particular to a parallel gripper robot picking method, device, equipment and product. BACKGROUND

[0002] In the fields of industrial automation, warehouse management, and logistics sorting, it is a core task for robots to pick and place objects in a cluttered environment, and the automation level directly affects production efficiency and management costs. However, robot picking in a cluttered environment faces multiple challenges: visual perception difficulties caused by object stacking and interlacing, collision risks between picking poses and the surrounding environment, and the influence of objects of different shapes and materials on picking stability, etc. These problems make it difficult for robots to achieve efficient and stable autonomous picking operations.

[0003] In the prior art, robot picking systems usually rely on complex sensor combinations (such as the combination of visual sensors and force sensors) or special mechanical structures, which not only increase system costs and integration difficulties, but also limit their application in low-cost scenarios. At the same time, traditional picking pose planning methods are often designed for single objects or structured environments, and in cluttered scenes with dense objects and various shapes, they are prone to problems such as unreasonable pose planning and weak collision avoidance capabilities, resulting in low picking success rates and poor operation efficiency.

[0004] Therefore, there is an urgent need for a new parallel gripper robot picking method. SUMMARY

[0005] In view of the above problems, the embodiments of the present application provide a parallel gripper robot picking method, device, electronic equipment and readable storage medium, so as to overcome the above problems or at least partially solve the above problems.

[0006] In a first aspect, the embodiments of the present application provide a parallel gripper robot picking method, which comprises:

[0007] Obtaining a point cloud image of a target scene, the target scene including at least two objects;

[0008] Based on the point cloud image and a pre-trained picking pose prediction model, obtaining a first picking pose for each object in the target scene;

[0009] Determining a second picking pose for a target object and a score of each second picking pose from the first picking pose of each object;

[0010] adjust the second picking posture in sequence from high to low according to the scores of the second picking postures until it is determined that the gripper of the parallel gripper robot does not have overlapping parts with the point cloud image in the third picking posture after the adjustment of a certain second picking posture; the strategy for adjusting the second picking posture comprises a storage box bottom collision avoidance strategy, a center correction strategy, an article collision avoidance strategy, and a storage box collision avoidance strategy;

[0011] controlling the parallel gripper robot to pick the target article from the target scene according to the third picking posture.

[0012] Optionally, in the case where the strategy for adjusting the second picking posture is the storage box bottom collision avoidance strategy, the adjustment of the second picking posture comprises:

[0013] moving the second picking posture upward by a first preset distance;

[0014] rotating the second picking posture after the movement around a Z axis as a rotation axis until the second picking posture after the rotation does not have overlapping parts with the point cloud image or reaches a maximum rotation angle, to obtain a third picking posture, wherein the Z axis is perpendicular to a plane formed by two grippers of the parallel gripper corresponding to the second picking posture.

[0015] Optionally, in the case where the strategy for adjusting the second picking posture is the center correction strategy, the adjustment of the second picking posture comprises:

[0016] extracting the point cloud image of the target article, and calculating the center of the point cloud image of the target article;

[0017] moving the picking center of the second picking posture to the center of the point cloud image to obtain a third picking posture.

[0018] Optionally, in the case where the strategy for adjusting the second picking posture is the article collision avoidance strategy, the adjustment of the second picking posture comprises:

[0019] in the case where the gripper base corresponding to the second picking posture has overlapping parts with the point cloud image, moving the second picking posture upward by a second preset distance;

[0020] in the case where the gripper finger corresponding to the second picking posture has overlapping parts with the point cloud image,

[0021] adjusting the second picking posture according to at least one of the following:

[0022] enlarging the second picking posture by a preset ratio;

[0023] The second picking posture is rotated around an X axis, which is parallel to a vertical direction of the parallel gripper corresponding to the second picking posture.

[0024] The second picking posture is moved upward by a third preset distance.

[0025] Optionally, in a case where the strategy for adjusting the second picking posture is a storage box collision avoidance strategy, the adjusting of the second picking posture comprises:

[0026] According to geometric parameters of the storage box, the gripper of the robot, and the mechanical arm of the robot, it is determined whether there is a collision between the mechanical arm and the storage box in the second picking posture;

[0027] In a case where it is determined that there is a collision, the second picking posture is rotated around a Y axis until it is determined that there is no collision between the mechanical arm and the storage box in the adjusted third picking posture, or a maximum rotation angle is reached, the Y axis being located in a plane formed by two grippers of the parallel gripper corresponding to the second picking posture and being perpendicular to a vertical direction of the parallel gripper corresponding to the second picking posture.

[0028] Optionally, after the parallel gripper robot is controlled to pick the target object from the target scene according to the third picking posture, the method further comprises:

[0029] The gripper opening width of the parallel gripper robot is determined;

[0030] In a case where the gripper opening width is greater than or equal to a first threshold value, it is determined that a picking result corresponding to the third picking posture is picking success;

[0031] In a case where the gripper opening width is less than the first threshold value, it is determined that the picking result corresponding to the third picking posture is picking failure.

[0032] In a second aspect, an embodiment of the present application provides a parallel gripper robot picking device, the device comprising:

[0033] An acquisition module is configured to acquire a point cloud image of a target scene, the target scene comprising at least two objects;

[0034] A posture prediction module is configured to obtain a first picking posture for each object in the target scene based on the point cloud image and a pre-trained picking posture prediction model.

[0035] A scoring module is configured to determine a second picking posture for a target object and a score of each second picking posture from the first picking postures of the objects.

[0036] The adjusting module is configured to sequentially adjust each second picking posture in descending order of the scores of the second picking postures until a third picking posture is determined, in which the gripper of the parallel gripper robot does not overlap with the point cloud image, and the third picking posture is adjusted from the second picking posture according to a strategy, the strategy including moving the posture upward, moving the picking center of the posture to the point cloud center of the target object, an object collision avoidance strategy, and a storage box collision avoidance strategy.

[0037] The picking module is configured to control the parallel gripper robot to pick the target object from the target scene according to the third picking posture.

[0038] In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory, and the processor executes the computer program to implement the parallel gripper robot picking method according to any one of the above aspects.

[0039] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the parallel gripper robot picking method according to the first aspect of the present application.

[0040] In a fifth aspect, a computer program product is provided, which includes computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the parallel gripper robot picking method according to the first aspect of the present application.

[0041] Specific beneficial effects are as follows:

[0042] In the embodiment of the present application, the point cloud image of the target cluttered scene (including at least two objects) is obtained, the point cloud image is input into the trained picking posture prediction model, and a plurality of first picking postures output by the picking posture prediction model are obtained. The picking posture prediction model is trained based on a sample scene training set and a gripper picking motion equation of the parallel gripper robot. Further, a plurality of second picking postures for the target object and scores are determined. Each second picking posture is adjusted in descending order of the scores of the second picking postures. The adjustment strategy includes a storage box bottom collision avoidance strategy, a center correction strategy, an object collision avoidance strategy, and a storage box collision avoidance strategy. Thus, a third picking posture is obtained, in which the gripper of the parallel gripper robot does not overlap with the point cloud image. Based on the third picking posture, the parallel gripper robot is controlled to pick objects in the target cluttered scene. In the embodiment of the present application, the scene features of the target cluttered scene can be accurately expressed by the point cloud image in the visual feedback method, and the parallel gripper robot can adapt to complex cluttered environments and accurately pick objects through adjustment of the candidate picking posture. The picking accuracy and efficiency of the robot in picking objects in the cluttered environment are improved. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating a parallel clamping robot picking method provided in an embodiment of the present invention;

[0045] Figure 2 This is an exemplary representation of the picking posture in a parallel clamping robot picking method provided in an embodiment of the present invention;

[0046] Figure 3 This is a logic block diagram of a parallel clamping robot picking device provided in an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0048] Exemplary embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0049] Reference Figure 1 , Figure 1 This is a flowchart illustrating a parallel gripper robot picking method according to an embodiment of the present invention. The method includes:

[0050] S101, acquire a point cloud image of the target scene, wherein the target scene includes at least two items.

[0051] In this embodiment of the invention, the target scene can be a cluttered scene, specifically a real-world scene containing many items with an irregular distribution. Point cloud images of the target scene can be obtained using methods such as LiDAR imaging, structured light scanning, and computer vision technology. The target scene includes the target item, as well as other objects that interfere with the target item.

[0052] In the embodiment of the present application, the point cloud image (Point Cloud) refers to a dataset for describing the shape of the surface of an object or a scene by three-dimensional coordinate points, which is an important form of three-dimensional information in the fields of three-dimensional scanning, computer vision, robot perception, etc. Unlike traditional two-dimensional images (such as photos) which only contain plane pixel information, each point in the point cloud image has accurate three-dimensional spatial coordinates (X, Y, Z), and some points may also carry additional information such as color (RGB), reflectivity, normal vector, etc.

[0053] In S102, based on the point cloud image and the pre-trained picking pose prediction model, a first picking pose for each object in the target scene is obtained.

[0054] In the embodiment of the present application, the picking pose prediction model is trained based on the sample scene training set and the clamping motion equation of the parallel clamping robot.

[0055] In the embodiment of the present application, the pre-trained picking pose prediction model can be a picking pose prediction model.

[0056] GraspNet is a high-efficiency neural network model for real-time detection of picking by low-power devices, which is derived from an open source project "GraspNet". In the project, there is a dataset "GraspNet-1Billion" and a picking algorithm "AnyGrasp". The dataset "GraspNet-1Billion" is a large-scale benchmark dataset, which contains 97,280 RGB-D images and more than 1 billion picking poses. These images are collected from 190 cluttered scenes, each scene contains multiple target picking objects, and accurate 6D poses and dense picking pose annotations are provided. The dataset can be used as training data for the picking pose prediction model; the picking algorithm "AnyGrasp" is a picking solution of the GraspNet project, aiming to achieve human-level picking ability, suitable for any object in a cluttered scene, whether it is a rigid or deformable object.

[0057] In the embodiment of the present application, the dataset "GraspNet-1Billion" or part of the data in the dataset can be used as a sample cluttered scene training set to train the picking pose prediction model. In the output layer of the picking pose prediction model, the picking motion equation of the parallel clamping robot when clamping the picking object can be used as the activation function of the output layer (i.e. the output layer can be a linear layer). In this way, a trained picking pose prediction model can be obtained. Then, the point cloud image can be input into the trained picking pose prediction model to obtain multiple first picking poses output by the picking pose prediction model. Obviously, the above-mentioned multiple first picking poses are all matched with the parallel clamping robot and can be correctly executed by the parallel clamping robot.

[0058] S103, determining a second picking pose for the target object and a score of each second picking pose from the first picking pose of each object.

[0059] In the embodiment of the present application, the picking pose prediction model can output a plurality of first picking poses for a plurality of objects included in the target scene, and the plurality of first picking poses include a plurality of picking poses for different objects. When determining the target object that needs to be picked currently, the picking pose for the target object, i.e., the second picking pose, can be determined from the plurality of first picking poses.

[0060] In the embodiment of the present application, when the picking pose prediction model outputs the first picking pose, it can also output the pose score corresponding to each first picking pose. The pose score is used to evaluate the degree of adaptation between the first picking pose and the corresponding object. The higher the pose score, the higher the degree of adaptation between the first picking pose and the corresponding object, and the higher the probability of successfully picking the object by the robot.

[0061] Therefore, the point cloud image can be input into the trained picking pose prediction model, so that the pose score corresponding to each first picking pose output by the picking pose prediction model can be obtained. If the first picking pose is displayed in the form of image annotation, the pose score corresponding to the first picking pose can also be displayed in the form of image annotation, and the color of the pose score corresponding to the first picking pose can be the same as the color of the first picking pose, so as to reflect the corresponding relationship between the first picking pose and the pose score.

[0062] In the embodiment of the present application, based on the output result of the picking pose prediction model, the score of each second picking pose can be obtained on the basis of the determination of the second picking pose.

[0063] S104, adjusting each second picking pose in turn according to the score of each second picking pose from high to low, until the third picking pose after the adjustment of a certain second picking pose is determined, and the gripper of the parallel gripper robot does not have overlapping parts with the point cloud image; the strategy for adjusting the second picking pose includes a collision avoidance strategy of the bottom of the storage box, a center correction strategy, an object collision avoidance strategy, and a collision avoidance strategy of the storage box.

[0064] In the embodiment of the present application, considering that the picking pose of the parallel gripper robot needs to meet certain conditions (the gripper of the parallel gripper robot does not have overlapping parts with the point cloud image) to successfully pick the object when picking the object, a plurality of picking poses can be evaluated and adjusted to obtain a picking pose that passes the evaluation.

[0065] In the embodiment of the present application, in order to enable the parallel gripper robot to adapt to a complex and cluttered environment and perform accurate picking, each second picking pose needs to be adjusted.

[0066] In the embodiment of the present application, each second picking pose can be adjusted and evaluated from high to low according to the scores of each second picking pose, so as to obtain a picking pose that can make the gripper of the parallel gripper robot not coincide with the point cloud image, so as to avoid the gripper robot from colliding with the object or the object box during picking.

[0067] In the embodiment of the present application, the evaluation criteria include whether the gripper of the parallel gripper robot coincides with the target object, whether the gripper collides with the target object when rotating, whether the distance between the gripper base and the target object meets the picking condition, etc. In the case that the gripper does not coincide with the target object, the gripper collides with the target object when rotating, and the distance between the gripper base and the target object meets the picking condition, it can be determined that the evaluation is passed.

[0068] In the embodiment of the present application, the parallel gripper robot can include a mechanical arm, and a gripper can be loaded at the free end of the mechanical arm for picking objects. The parallel gripper robot can control the gripper to move in space, and the moving range of the gripper is a spherical space with the fixed end of the mechanical arm as the center and the maximum distance between the fixed end of the mechanical arm and the free end of the mechanical arm as the radius. The fixed end of the mechanical arm can be a mechanical arm mounting point in the parallel gripper robot, and the fixed end can move with the movement of the parallel gripper robot.

[0069] In the embodiment of the present application, the object box bottom collision avoidance strategy refers to adjusting the position and angle of the second picking pose to avoid the pose being too low, which may cause the gripper of the parallel gripper robot to collide with the bottom of the object box.

[0070] In the embodiment of the present application, the center correction strategy refers to adjusting the picking center position of the second picking pose to correct the picking pose in which the picking center is located at the edge of the object.

[0071] In the embodiment of the present application, the object collision avoidance strategy refers to adjusting the position and angle of the second picking pose to avoid the gripper (including the gripper base and the gripper fingers) of the parallel gripper robot colliding with any object in the target scene.

[0072] In the embodiment of the present application, the object box collision avoidance strategy refers to adjusting the position and angle of the second picking pose to avoid the mechanical arm of the parallel gripper robot colliding with the object box.

[0073] In the embodiments of the present application, the second picking posture can be iteratively adjusted starting from the posture with the highest score. If a feasible posture (the jaws of the parallel-jaw robot in the third picking posture do not overlap with the point cloud image) is obtained after adjustment, the third picking posture (i.e., the feasible posture) can be determined and immediately executed. If a feasible posture is not obtained, the current second picking posture is discarded, and the next posture in the score list is processed, and the process is repeated until a feasible posture is found or all postures are evaluated.

[0074] In the embodiments of the present application, when the strategy for adjusting the second picking posture is the parcel box bottom collision avoidance strategy, adjusting the second picking posture comprises:

[0075] moving the second picking posture upward by a first preset distance;

[0076] rotating the moved second picking posture around the Z axis as the rotation axis until the rotated second picking posture does not overlap with the point cloud image or reaches a maximum rotation angle, to obtain a third picking posture, wherein the Z axis is perpendicular to the plane formed by the two jaws of the parallel jaw corresponding to the second picking posture.

[0077] Specifically, in the embodiments of the present application, the picking posture can be represented as the shape of the jaws of the parallel-jaw robot, that is, each picking posture corresponds to the two jaw fingers and the jaw base of the parallel-jaw robot. Specifically, Figure 2 An exemplary representation of a picking posture is shown. In the representation, the Z axis is perpendicular to the plane formed by the two jaws of the parallel jaw corresponding to the picking posture, the X axis is parallel to the vertical direction of the parallel jaw corresponding to the picking posture, and the Y axis is located in the plane formed by the two jaws of the parallel jaw corresponding to the picking posture and is perpendicular to the vertical direction of the parallel jaw corresponding to the picking posture.

[0078] When the picking posture adopted by the robot is too low, the height of the jaws is close to or lower than the bottom of the parcel box, or the motion trajectory of the jaws overlaps with the bottom of the parcel box. During the process of picking the items in the parcel box, if the posture control is improper, the jaws may scratch, squeeze or even get stuck on the bottom of the box. Therefore, the parcel box bottom collision avoidance strategy is proposed in the embodiments of the present application. Specifically, moving the second picking posture upward by a first preset distance means moving the picking posture upward as a whole to increase the vertical distance between the jaws and the bottom of the parcel box. When the second picking posture is too low, during the process of the robot working according to the second picking posture, the jaws may be close to or even contact the bottom of the box. After moving upward, the distance between the two can be pulled apart in space, leaving a safety gap for subsequent actions. Similar to when a person picks up something, if the hand is too low and may touch the table, lifting the hand (moving the posture upward) can avoid contact.

[0079] In the embodiments of the present application, the plane of the clamp refers to the plane in which the two clamps of the parallel clamp are located (usually the plane in which the opening / closing action of the clamp is located).

[0080] In the embodiments of the present application, the first preset distance can be set by a person according to a certain proportion of the length of the clamp fingers.

[0081] In the embodiments of the present application, the second picking posture after movement is rotated around the Z axis, which can adjust the orientation of the clamp. In actual application, even if the posture is moved up, if the angle of the clamp is not suitable (for example, the tilt direction is still directed at the bottom of the box), it may still collide during movement or operation. After rotation, the spatial orientation of the clamp can be changed to avoid the area of the bottom of the box.

[0082] In the embodiments of the present application, in the case where the strategy for adjusting the second picking posture is the center correction strategy, the adjusting of the second picking posture comprises:

[0083] extracting the point cloud image of the target object, and calculating the center of the point cloud image of the target object;

[0084] moving the picking center of the second picking posture to the center of the point cloud image to obtain a third picking posture.

[0085] In the embodiments of the present application, the picking posture refers to the spatial posture of the parallel clamp robot when preparing to pick an object, including the position, angle, height, etc. of the clamp (including the clamp fingers and the clamp base) of the parallel clamp robot, and the core target is to enable the clamp to stably and accurately contact and pick the object.

[0086] After determining the preliminary picking posture, there may be a deviation in the picking center corresponding to the current picking posture (especially when the point picking center is located at the edge of the object), which needs to be corrected through subsequent steps.

[0087] Point cloud is a collection of massive three-dimensional coordinate points obtained by scanning an object with a three-dimensional sensor (such as a laser radar or a depth camera), each point containing X, Y, Z coordinates, which together constitute the surface shape of the object (similar to the three-dimensional profile of "digital twin"). For example, scanning an apple, the point cloud is the three-dimensional position data of the countless points on the surface of the apple. Extracting the point cloud of the target object focuses on the actual surface information of the object in this area, and the point cloud image of the target object directly reflects the true shape of the object, avoiding interference from the point cloud of the background (such as the desktop or other objects).

[0088] The center of the point cloud image of the target object calculated is closer to the physical center of the object (rather than the initially set theoretical center), because the point cloud directly comes from the surface of the object and can reflect the true shape of the object (such as whether there are depressions or protrusions).

[0089] The picking center originally set in the second picking posture (may have deviation) is updated to the center of the point cloud image calculated in the previous step, and the picking center can be adjusted by coordinates to coincide with the center of the actual surface of the object. The initial picking center that may fall on the edge of the object (instead of the center area) due to identification errors (such as visual algorithm misjudgment, object surface reflection / occlusion) is corrected.

[0090] For example, if the article is a cup, the initial picking center may misjudge the edge of the cup mouth (instead of the center of the cup mouth), causing the tongs to slip when picking; if the object is an irregularly shaped part, the initial center may fall on the protruding edge, causing uneven force when picking, and the object to fall off. After correction, the picking center is moved from the edge to the substantial center area of the object, making the contact between the tongs and the article more stable, reducing the risk of slipping and picking failure.

[0091] In the embodiment of the application, in the case that the strategy for adjusting the second picking posture is the article collision avoidance strategy, adjusting the second picking posture comprises:

[0092] In the case that the tongs base corresponding to the second picking posture overlaps with the point cloud image, the second picking posture is moved upward by a second preset distance;

[0093] In the case that the tongs fingers corresponding to the second picking posture overlap with the point cloud image,

[0094] Adjust the second picking posture according to at least one of the following:

[0095] Amplify the second picking posture by a preset ratio;

[0096] Rotate the second picking posture around the X axis as the rotation axis, and the X axis is parallel to the vertical direction of the parallel tongs corresponding to the second picking posture;

[0097] Move the second picking posture upward by a third preset distance.

[0098] In the embodiment of the application, whether the virtual model of the tongs (generated based on the parameters of the current second picking posture) and the point cloud image overlap in space is simulated by an algorithm. If there is overlap, it means that the tongs will collide with the object / environment in this posture (such as the tongs colliding with the surface of the article, the wall of the object box, etc.).

[0099] In the embodiment of the application, different methods are proposed to adjust the posture according to the collision. Specifically, the collision may occur at different parts of the tongs (tongs base or tongs fingers).

[0100] If the gripper base (such as the lower end surface of the entire gripper, not the finger part) overlaps with the point cloud, it usually means that the height of the gripper is too low, and the bottom is too close to the surface of the object (such as the gripper base colliding with the protruding part of the object). In this case, the adjustment strategy is to move the posture up by a second preset distance to increase the vertical distance between the gripper base and the point cloud (the surface of the object) and avoid collision in space.

[0101] In the embodiments of the present application, considering that most of the collisions of the gripper base are caused by "insufficient height", moving up is the most direct and effective solution, which does not require complex adjustment, thereby reducing the algorithm complexity and improving the adjustment efficiency.

[0102] If the two gripper fingers (components performing the picking action) of the gripper overlap with the point cloud image, it means that the gripper fingers will collide with the object in the current second picking posture (such as the gripper fingers inserting into the object, colliding with the side of the object / protrusion, or the space between the gripper fingers being too small to be stuck by the object).

[0103] In the embodiments of the present application, considering that the reasons for the collision of the gripper fingers with the object are more complex, such as insufficient finger opening angle, inappropriate angle, or slightly low height, a single strategy may not be effective, therefore, in this case, the embodiments of the present application propose to randomly perform different adjustment actions.

[0104] In the embodiments of the present application, actions can be randomly selected according to a preset probability until there is no collision or the maximum number of attempts is reached (to avoid infinite loop), and the purpose and logic of each action are as follows:

[0105] Amplifying the second picking posture by a preset ratio (for example, the probability is 0.45):

[0106] In the embodiments of the present application, amplifying the second picking posture by a preset ratio can expand the posture range, that is, the opening range of the two fingers of the gripper (that is, the distance between the gripper fingers). Thus, by increasing the opening angle of the gripper fingers and increasing the space between the gripper fingers, collisions with the side of the object / protrusion due to insufficient finger spacing can be avoided (for example, when picking a wider object, insufficient opening of the gripper fingers will collide with the two sides of the object).

[0107] In the embodiments of the present application, considering that the space between the gripper fingers being too small to be stuck by the object is a common collision reason, therefore, a higher probability can be set for the action of "amplifying the second picking posture by a preset ratio".

[0108] Rotating the second picking posture around the X-axis as the rotation axis (for example, the probability is 0.45):

[0109] In the embodiment of the present application, the X axis refers to the axis along the length direction of the clamp (for example, the direction from the root to the tip of the clamp finger is the X axis), and rotating around the X axis can adjust the pitch angle of the clamp (similar to rotating up and down on the wrist). Therefore, by changing the orientation of the finger, the finger can avoid overlapping with the object (for example, the object has a protrusion on the side, and after rotating, the finger approaches from the oblique upper side instead of the front, which can avoid the protrusion).

[0110] In the embodiment of the present application, considering that the angle deviation is another common cause of collision between the clamp and the object, the rotation adjustment can flexibly change the spatial orientation of the finger and adapt to the shape of the complex object, so a higher probability can be set for the action of "rotating the second picking posture around the X axis as the rotation axis".

[0111] Moving the second picking posture upward by a third preset distance (for example, probability 0.1)

[0112] In the embodiment of the present application, the collision between the clamp finger and the object can also be caused by insufficient height, so the processing of the collision with the clamp base can be similar, and the vertical overlap between the finger and the object can be reduced by increasing the height of the clamp (for example, there is a protrusion of the object below the finger, which can be avoided after moving upward).

[0113] In the embodiment of the present application, considering that the height of the clamp has been adjusted in the case where the clamp base and the point cloud image overlap, the collision of the finger is mostly a horizontal (range or angle) problem, and the height problem is less common, so a lower probability can be set for the action of "moving the second picking posture upward by a third preset distance" as a supplementary adjustment strategy.

[0114] In the embodiment of the present application, in the case where the strategy for adjusting the second picking posture is the collision avoidance strategy of the storage box, the adjustment of the second picking posture includes:

[0115] According to the geometric parameters of the storage box, the clamp of the robot, and the mechanical arm of the robot, it is determined whether there is a collision between the mechanical arm and the storage box in the second picking posture.

[0116] In the case where it is determined that there is a collision, the second picking posture is rotated around the Y axis until it is determined that there is no collision between the mechanical arm and the storage box in the adjusted third picking posture, or the maximum rotation angle is reached, the Y axis is located in the plane formed by the two clamps of the parallel clamp corresponding to the second picking posture, and is perpendicular to the vertical direction of the parallel clamp corresponding to the second picking posture.

[0117] In the embodiment of the present application, the geometric parameters of the storage box include length, width, height, shape (such as whether there is a protrusion, opening direction), spatial position (coordinates on the workbench), etc., which determine the area occupied by the storage box in space. The geometric parameters of the gripper include gripper finger length, width, thickness, opening angle, gripper base shape, etc. The geometric parameters of the robot arm include joint length, arm span, joint movement angle limit, etc., which reflect the overall movement trajectory and occupied range of the robot arm in space.

[0118] The geometric parameters of the storage box, the gripper of the robot, and the robot arm of the robot together constitute a "virtual geometric model" of the robot arm and the storage box, and whether a collision will occur can be determined by calculating whether there is an overlapping area (i.e., position intersection) in space between the two models.

[0119] Specifically, based on the geometric parameters of the storage box, the gripper of the robot, and the robot arm of the robot, the relative position relationship between the current second picking posture (position and angle) of the robot arm and the storage box can be simulated in a virtual space. If the geometric model of any part (such as the arm rod, joint, or gripper) of the robot arm overlaps with the geometric model of the storage box in space (i.e., both occupy the same three-dimensional space at the same time), it is determined that the storage box collides. For example, during movement of the robot arm, the arm rod may collide with the side of the storage box; when the gripper approaches the storage box, the gripper may overlap with the edge of the box opening; when the robot arm posture is too low, the bottom joint may interfere with the top of the storage box.

[0120] In the embodiment of the present application, when the storage box collision is detected, targeted adjustment can be taken: rotating around the Y-axis of the posture until no collision is detected or the maximum rotation angle is reached.

[0121] Rotating the second picking posture around the Y-axis can change the relative angle between the robot arm and the storage box, so that the geometric model of the robot arm and the geometric model of the storage box are "offset" in space, thereby eliminating the overlapping area and avoiding collision.

[0122] For example, if the arm rod of the robot arm collides with the side of the storage box due to angle problems, after rotating a certain angle around the Y-axis, the arm rod may move from the side of the storage box to an unobstructed area in front of or behind the storage box, avoiding collision.

[0123] In the embodiment of the present application, rotating around the Y-axis can change the height (Z-axis direction) and the front-back / left-right translation position (X, Y-axis coordinates) of the robot arm, and only through angle change can diversified adjustment be achieved to avoid collision, which is simple and efficient to operate and is suitable for solving side, edge, and other angle deviation caused collisions. Compared with other adjustment methods (such as up-down movement, expanding the range), rotating around the Y-axis is easier to control and can achieve diversified adjustment, quickly testing the collision possibility under different conditions.

[0124] In the embodiment of the present application, the XYZ axis of the picking posture is predefined, and in different adjustment strategies, the picking posture is controlled to rotate around different axes to adjust the angle of the picking posture, so as to avoid the robot colliding with the bottom of the storage box, the goods and the storage box in the picking posture.

[0125] S105, controlling the parallel jaw robot to pick the target goods from the target scene according to the third picking posture.

[0126] In the embodiment of the present application, when picking goods, the jaw of the parallel jaw robot can be controlled to move to the spatial position corresponding to the third picking posture, and then the third picking posture is executed to realize the picking of goods. The picking result can include picking success and picking failure. Wherein, the failure to pick goods and the falling of goods can be determined as picking failure, and the goods not falling can be determined as picking success.

[0127] In the embodiment of the present application, after controlling the parallel jaw robot to pick the target goods from the target scene according to the third picking posture, the method further comprises:

[0128] S201, determining the jaw opening width of the jaw of the parallel jaw robot.

[0129] S202, in the case that the jaw opening width is greater than or equal to a first threshold, determining that the picking result corresponding to the third picking posture is picking success; in the case that the jaw opening width is less than the first threshold, determining that the picking result corresponding to the third picking posture is picking failure.

[0130] In the embodiment of the present application, if the jaw opening width of the jaw is greater than or equal to the first threshold, it can be considered that the picking result corresponding to the third picking posture is picking success. Wherein, the first threshold can be set according to the distance between the fingers of the jaw when the jaw performs picking without picking goods. If the jaw opening width of the jaw is less than the first threshold, it can be determined that the picking result is picking failure.

[0131] In the embodiment of the present application, after the parallel jaw robot picks the target goods from the target scene according to the third picking posture, the jaw opening width of the jaw of the parallel jaw robot is determined, and the picking result is determined to be picking success or picking failure according to the size relationship between the jaw opening width of the jaw and the first threshold, which can improve the accuracy and reliability of the picking result of the robot picking to a certain extent.

[0132] In the case of a successful picking result, the parallel gripper robot can be controlled to move the target object corresponding to the third picking posture to the target area. This "picking-moving-placing" execution process is a complete robot picking process. After completing a placement, the parallel gripper robot can start the next "picking-moving-placing" process.

[0133] In the embodiment of the present application, the point cloud image of the target cluttered scene (including at least two objects) is obtained, the point cloud image is input into the trained picking posture prediction model, and a plurality of first picking postures output by the picking posture prediction model are obtained. The picking posture prediction model is trained based on a sample scene training set and a gripper picking motion equation of the parallel gripper robot, a plurality of second picking postures for the target object and scores are further determined, each second picking posture is adjusted in turn according to the scores of each second picking posture, the adjustment strategies include a collision avoidance strategy of the bottom of the storage box, a center correction strategy, an object collision avoidance strategy and a storage box collision avoidance strategy, so as to obtain a third picking posture in which the gripper of the parallel gripper robot does not overlap with the point cloud image. Based on the third picking posture, the parallel gripper robot is controlled to pick objects in the target cluttered scene, so that in the embodiment of the present application, the scene features of the target cluttered scene can be accurately expressed through the point cloud image in the visual feedback method, and the parallel gripper robot can adapt to complex cluttered environments and accurately pick objects through the adjustment of the candidate picking postures, thereby improving the picking accuracy and efficiency of the robot in the cluttered environment.

[0134] Reference Figure 3 , Figure 3 A logic block diagram of a parallel gripper robot picking device provided by the embodiment of the present application can include:

[0135] The acquisition module 301 is configured to acquire a point cloud image of a target scene, and the target scene includes at least two objects.

[0136] The posture prediction module 302 is configured to obtain a first picking posture for each object in the target scene based on the point cloud image and a pre-trained picking posture prediction model.

[0137] The scoring module 303 is configured to determine a second picking posture for a target object and a score of each second picking posture from the first picking postures of each object.

[0138] The adjusting module 304 is configured to sequentially adjust each second picking posture in a descending order of scores of the second picking postures, until a third picking posture after adjustment of a certain second picking posture is determined, in which the gripper of the parallel gripper robot does not overlap with the point cloud image; and the strategy for adjusting the second picking posture comprises: moving the posture upward, moving the picking center of the posture to the point cloud center of the target object, an object collision avoidance strategy, and a storage box collision avoidance strategy.

[0139] The picking module 305 is configured to control the parallel gripper robot to pick the target object from the target scene according to the third picking posture.

[0140] Optionally, in a case where the strategy for adjusting the second picking posture is the storage box bottom collision avoidance strategy, the adjusting module is configured to:

[0141] move the second picking posture upward by a first preset distance;

[0142] rotate the second picking posture after the movement around a Z axis as a rotation axis until the second picking posture after the rotation does not overlap with the point cloud image or reaches a maximum rotation angle, to obtain the third picking posture, wherein the Z axis is perpendicular to a plane formed by the two grippers of the parallel gripper corresponding to the second picking posture.

[0143] Optionally, in a case where the strategy for adjusting the second picking posture is the center correction strategy, the adjusting module is configured to:

[0144] extract the point cloud image of the target object, and calculate a center of the point cloud image of the target object;

[0145] move a picking center of the second picking posture to the center of the point cloud image to obtain the third picking posture.

[0146] Optionally, in a case where the strategy for adjusting the second picking posture is the object collision avoidance strategy, the adjusting module is configured to:

[0147] in a case where a gripper base corresponding to the second picking posture overlaps with the point cloud image, move the second picking posture upward by a second preset distance;

[0148] in a case where a gripper finger corresponding to the second picking posture overlaps with the point cloud image,

[0149] adjust the second picking posture according to at least one of the following:

[0150] enlarge the second picking posture by a preset ratio;

[0151] The second picking posture is rotated around an X axis as a rotation axis, and the X axis is parallel to a vertical direction of the parallel gripper corresponding to the second picking posture.

[0152] The second picking posture is moved upward by a third preset distance.

[0153] Optionally, in a case where the strategy for adjusting the second picking posture is a storage box collision avoidance strategy, the adjusting module is configured to:

[0154] According to geometric parameters of the storage box, the gripper of the robot, and the mechanical arm of the robot, it is determined whether there is a collision between the mechanical arm and the storage box in the second picking posture.

[0155] In a case where it is determined that there is a collision, the second picking posture is rotated around a Y axis until it is determined that there is no collision between the mechanical arm and the storage box in the adjusted third picking posture, or a maximum rotation angle is reached, the Y axis is located in a plane formed by two grippers of the parallel gripper corresponding to the second picking posture and is perpendicular to a vertical direction of the parallel gripper corresponding to the second picking posture.

[0156] Optionally, the device further comprises a result detection module configured to:

[0157] The mouth width of the gripper of the parallel gripper robot is determined.

[0158] In a case where the mouth width is greater than or equal to a first threshold value, it is determined that a picking result corresponding to the third picking posture is picking success.

[0159] In a case where the mouth width is less than the first threshold value, it is determined that the picking result corresponding to the third picking posture is picking failure.

[0160] The parallel-jaw robot picking device in the embodiment of the present application can be an electronic device, or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices other than the terminal. For example, the electronic device can be a GPU BOX, a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a Mobile Internet Device (MID), an augmented reality (AR) / virtual reality (VR) device, a robot, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), and can also be a server, a Network Attached Storage (NAS), a personal computer (PC), a television (TV), a teller machine, or a self-service machine, and the like, and the embodiment of the present application is not limited in this regard.

[0161] The parallel-jaw robot picking device in the embodiment of the present application can be a device with an operating system. The operating system can be an Android operating system, can be a Linux or Windows operating system, or can be other possible operating systems, and the embodiment of the present application is not limited in this regard.

[0162] The parallel-jaw robot picking device provided in the embodiment of the present application can implement Figure 1 The processes implemented by the method embodiment are not repeated here to avoid repetition.

[0163] The embodiment of the present application provides an electronic device, referring to Figure 4 The electronic device 40 includes a processor 401, a memory 402, and a computer program 4021 stored in the memory 402 and executable on the processor 401, and the processor 401 implements the parallel-jaw robot picking method of the foregoing embodiment when executing the program.

[0164] The embodiment of the present application further provides a computer-readable storage medium having a computer program / instruction stored thereon, and the computer program / instruction is executed by a processor to implement the steps in the parallel-jaw robot picking method disclosed in the embodiment of the present application.

[0165] The embodiment of the present application further provides a computer program product, which, when running on an electronic device, causes a processor to implement the steps in the parallel gripper robot picking method disclosed by the embodiment of the present application.

[0166] Each of the embodiments in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other.

[0167] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices, electronic devices and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal equipment to produce a machine, so that the instructions executed by the computer or other programmable data processing terminal equipment produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The functions specified in one flow or multiple flows and / or blocks

[0168] These computer program instructions can also be stored in a computer readable memory capable of guiding the computer or other programmable data processing terminal equipment to work in a specific manner, so that the instructions stored in the computer readable memory produce a product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The functions specified in one flow or multiple flows and / or blocks

[0169] These computer program instructions can also be loaded into the computer or other programmable data processing terminal equipment, so that a series of operation steps are performed on the computer or other programmable terminal equipment to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal equipment provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The functions specified in one flow or multiple flows and / or blocks

[0170] Although the preferred embodiments of the embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to the embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present application.

[0171] Finally, it is to be understood that the phraseology or terminology such as "first" and "second" etc. used herein is merely intended to distinguish one entity or operation from another without necessarily requiring or implying any actual such relationship or order between such entities or operations. Moreover, the use of the term "including", "containing" or any other variant thereof is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or even inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the recited element.

[0172] The parallel gripper robot picking method provided by the application is described in detail above, and the principles and implementation manners of the application are described by using specific examples in this paper. The above description of the examples is only used to help understand the method of the application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges will be changed according to the idea of the application. In summary, the content of this specification should not be understood as a limitation of the application.

Claims

1. A parallel-jawed robotic picking method, characterized by, The method comprises: acquiring a point cloud image of a target scene, the target scene comprising at least two articles; based on the point cloud image and a pre-trained picking posture prediction model, obtaining a first picking posture for each article in the target scene; determining a second picking posture for a target article and a score of each second picking posture from the first picking posture of each article; adjusting each second picking posture in turn according to the score of each second picking posture from high to low until determining that the gripper of the parallel gripper robot does not have an overlapping part with the point cloud image in a third picking posture after adjustment of a certain second picking posture; the strategy for adjusting the second picking posture comprises a storage box bottom collision avoidance strategy, a center correction strategy, an article collision avoidance strategy and a storage box collision avoidance strategy; through the adjustment of the second picking posture, the parallel gripper robot can adapt to a complex cluttered environment and perform accurate picking, and objects of different shapes and materials are stacked and interlaced in the cluttered environment; controlling the parallel gripper robot to pick the target article from the target scene according to the third picking posture; the article collision avoidance strategy refers to adjusting the position and angle of the second picking posture to avoid collision between the gripper base and the gripper fingers of the parallel gripper robot and any article in the target scene; in the case that the strategy for adjusting the second picking posture is the article collision avoidance strategy, adjusting the second picking posture comprises: in the case that the gripper base corresponding to the second picking posture has an overlap with the point cloud image, moving the second picking posture upward by a second preset distance; in the case that the gripper fingers corresponding to the second picking posture have an overlap with the point cloud image, adjusting the second picking posture according to at least one of the following until there is no collision or the maximum number of attempts is reached: based on a first probability, selecting a first action, the first action being: amplifying the second picking posture by a preset ratio to increase the opening angle of the gripper fingers; based on a second probability, selecting a second action, the second action being: rotating the second picking posture around the X axis as the rotation axis to adjust the pitch angle of the gripper, the X axis being parallel to the vertical direction of the parallel gripper corresponding to the second picking posture; based on a third probability, selecting a third action, the third action being: moving the second picking posture upward by a third preset distance to increase the height of the gripper and reduce the vertical overlap of the gripper fingers and the object; the third probability is lower than the first probability and lower than the second probability; in the case that the strategy for adjusting the second picking posture is the storage box collision avoidance strategy, adjusting the second picking posture comprises: Determine whether there is a collision between the robot arm and the storage box in the second picking pose according to the geometric parameters of the storage box, the gripper of the robot, and the mechanical arm of the robot; the geometric parameters of the storage box include length, width, height, shape, and spatial position, which determine the area occupied by the storage box in space; the geometric parameters of the gripper include gripper finger length, width, thickness, opening angle, and shape of the gripper base; and the geometric parameters of the mechanical arm include joint length, arm span range, and joint activity angle limit, which reflect the overall motion trajectory and occupied range of the mechanical arm in space; In the case where it is determined that there is a collision, rotate the second picking pose around the Y axis to change the relative angle between the mechanical arm and the storage box, so that the geometric model of the mechanical arm and the geometric model of the storage box are "offset" in space, thereby eliminating the overlapping area and avoiding collision, until it is determined that there is no collision between the mechanical arm and the storage box in the adjusted third picking pose, or the maximum rotation angle is reached, the Y axis being located in the plane formed by the two grippers of the parallel gripper corresponding to the second picking pose and being perpendicular to the vertical direction of the parallel gripper corresponding to the second picking pose.

2. The method of claim 1, wherein, In the case where the strategy for adjusting the second picking pose is the storage box bottom collision avoidance strategy, adjusting the second picking pose includes: Moving the second picking pose upward by a first preset distance; Rotating the moved second picking pose around the Z axis until the rotated second picking pose does not overlap with the point cloud image or the maximum rotation angle is reached, to obtain a third picking pose, the Z axis being perpendicular to the plane formed by the two grippers of the parallel gripper corresponding to the second picking pose.

3. The method of claim 1, wherein, In the case where the strategy for adjusting the second picking pose is the center correction strategy, adjusting the second picking pose includes: Extracting the point cloud image of the target object and calculating the center of the point cloud image of the target object; Moving the picking center of the second picking pose to the center of the point cloud image to obtain a third picking pose.

4. The method according to any one of claims 1 to 3, characterized in that, After controlling the parallel gripper robot to pick the target object from the target scene according to the third picking pose, the method further includes: Determining the opening width of the gripper of the parallel gripper robot; In the case where the opening width is greater than or equal to a first threshold, determining that the picking result corresponding to the third picking pose is picking success; In the case where the opening width is less than the first threshold, determining that the picking result corresponding to the third picking pose is picking failure.

5. A parallel-jawed robotic picking device, characterized in that The device includes: An acquisition module configured to acquire a point cloud image of a target scene, the target scene including at least two objects; A pose prediction module configured to obtain a first picking pose for each object in the target scene based on the point cloud image and a pre-trained picking pose prediction model; A scoring module configured to determine a second picking pose for a target object and a score of each second picking pose from the first picking poses of the objects. The adjusting module is configured to sequentially adjust each second picking pose in descending order of scores of the second picking poses until a third picking pose after adjustment of a certain second picking pose is determined, in which the gripper of the parallel gripper robot does not overlap with the point cloud image; the strategy for adjusting the second picking pose includes: moving the pose upward, moving the picking center of the pose to the point cloud center of the target object, an object collision avoidance strategy, and a storage box collision avoidance strategy; through the adjustment of the second picking pose, the parallel gripper robot can adapt to a complex cluttered environment and perform accurate picking, and objects of different shapes and materials are stacked and interlaced in the cluttered environment; The picking module is configured to control the parallel gripper robot to pick the target object from the target scene according to the third picking pose. The object collision avoidance strategy refers to adjusting the position and angle of the second picking pose to avoid collision between the gripper base and the gripper fingers of the parallel gripper robot and any object in the target scene; in the case where the strategy for adjusting the second picking pose is the object collision avoidance strategy, adjusting the second picking pose includes: In the case where the gripper base corresponding to the second picking pose overlaps with the point cloud image, moving the second picking pose upward by a second preset distance; In the case where the gripper finger corresponding to the second picking pose overlaps with the point cloud image, adjusting the second picking pose according to at least one of the following until there is no collision or the maximum number of attempts is reached: selecting a first action based on a first probability, the first action being: amplifying the second picking pose by a preset ratio to increase the opening angle of the gripper fingers; selecting a second action based on a second probability, the second action being: rotating the second picking pose about the X axis as the rotation axis to adjust the pitch angle of the gripper, the X axis being parallel to the vertical direction of the parallel gripper corresponding to the second picking pose; selecting a third action based on a third probability, the third action being: moving the second picking pose upward by a third preset distance to increase the height of the gripper and reduce the vertical overlap between the gripper fingers and the object; the third probability is lower than the first probability and lower than the second probability; In the case where the strategy for adjusting the second picking pose is the storage box collision avoidance strategy, adjusting the second picking pose includes: determining whether there is a collision between the mechanical arm and the storage box in the second picking pose according to the geometric parameters of the storage box, the gripper of the robot, and the mechanical arm of the robot; the geometric parameters of the storage box include: length, width, height, shape, and spatial position, which determine the area occupied by the storage box in space; the geometric parameters of the gripper include: gripper finger length, width, thickness, opening angle, and shape of the gripper base; the geometric parameters of the mechanical arm include: joint length, arm span range, and joint activity angle limit, which reflect the overall motion trajectory and occupied range of the mechanical arm in space; In the case of determining the existence of collision, the second picking posture is rotated around the Y axis, the relative angle between the mechanical arm and the storage box is changed, the geometric model of the mechanical arm and the geometric model of the storage box are "offset" in space, so as to eliminate the overlapping area and avoid collision, until it is determined that there is no collision between the mechanical arm and the storage box in the adjusted third picking posture, or the maximum rotation angle is reached, the Y axis is located in the plane formed by the two clamps of the parallel clamp corresponding to the second picking posture and is perpendicular to the vertical direction of the parallel clamp corresponding to the second picking posture.

6. An electronic device, comprising: A computer program product comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program to implement the parallel clamp robot picking method of any one of claims 1 to 4.

7. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the parallel clamp robot picking method of any one of claims 1 to 4.

8. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instructions are executed by the processor to implement the steps in the parallel clamp robot picking method of any one of claims 1 to 4.

Images