Autonomous obstacle avoidance grabbing system and method

Abstract

The invention relates to the technical field of mechanical arm obstacle avoidance grabbing, and discloses an automatic obstacle avoidance grabbing system and method. The system comprises a mechanical arm, a stereo camera, a flexible grabbing unit and a control unit, the mechanical arm is arranged on a table top, the stereo camera is arranged on the mechanical arm, the flexible grabbing unit is arranged at the tail end of the mechanical arm, and the stereo camera is used for obtaining point cloud images and plane images of a target object and an obstacle; the control unit is used for obtaining point cloud data of the target object and the obstacle according to the point cloud images and the plane images of the target object and the obstacle, and processing the point cloud data of the target object and the obstacle to obtain the position of a grabbing point; envelope box description is carried out on the mechanical arm, the stereo camera, the flexible grabbing unit and the obstacle so as to carry out self-collision detection and obstacle collision detection, the path from the initial position to the target position is calculated under the condition that collision does not exist, and the mechanical arm is controlled to grab the target object according to the calculated path.

Description

Technical Field

[0001] The present invention relates to the technical field of robot arm obstacle avoidance and grasping, and in particular to an autonomous obstacle avoidance and grasping system and method. Background Art

[0002] Robotic grasping based on robotic arms and vision guidance is widely used in the industrial field. The general solution only uses vision to identify the target object to be grasped and directly guides the robotic arm to grasp it. However, there are generally various physical obstacles in the working space, and directly guiding the grasping has the risk of encountering physical obstacles. Summary of the invention

[0003] The present invention provides an autonomous obstacle avoidance and grasping system and method, which can solve the technical problems in the prior art.

[0004] The present invention provides an autonomous obstacle avoidance grasping system, wherein the system comprises a mechanical arm, a stereo camera, a flexible grasping unit and a control unit, wherein the mechanical arm is arranged on a table, the stereo camera is arranged on the mechanical arm, and the flexible grasping unit is arranged at the end of the mechanical arm.

[0005] The stereo camera is used to obtain point cloud images and plane images of the target object and obstacles;

[0006] The control unit is used to obtain point cloud data of the target object and the obstacle based on the point cloud image and the plane image of the target object and the obstacle, process the point cloud data of the target object and the obstacle to obtain the position of the grasping point, describe the robotic arm, the stereo camera, the flexible grasping unit and the obstacle with an envelope box to perform self-collision detection and obstacle collision detection, and calculate the path from the starting position to the target position in the absence of a collision, and control the robotic arm to grasp the target object according to the calculated path.

[0007] Preferably, obtaining point cloud data of the target object and the obstacle according to the point cloud image and the plane image of the target object and the obstacle includes:

[0008] Conditional filtering is performed on the point cloud images and plane images of the target and obstacles, plane segmentation is performed on the filtered images to remove the platform surface, and Euclidean distance clustering segmentation is performed on the images after the platform surface is removed to obtain point cloud data of the target and obstacles.

[0009] Preferably, obtaining point cloud data of the target object and the obstacle according to the point cloud image and the plane image of the target object and the obstacle includes:

[0010] Based on deep learning target detection, the target and obstacle are located according to their point cloud images and plane images, and the target and obstacle areas are segmented to obtain the point cloud data of the target and obstacle.

[0011] Preferably, processing the point cloud data of the target object and the obstacle to obtain the position of the grasping point includes:

[0012] The point cloud data of the target object is fitted or registered, and the point cloud data of the obstacle is enveloping to obtain the position of the grasping point.

[0013] Preferably, the self-collision detection includes collision detection between the arm bars of the robotic arm and collision detection between the flexible grasping unit and the stereo camera; and the obstacle collision detection includes collision detection between the robotic arm and the obstacle.

[0014] Preferably, performing an envelope box description on the robotic arm, the stereo camera, the flexible grasping unit and the obstacle to perform self-collision detection and obstacle collision detection comprises:

[0015] The robot arm is described by a cylindrical envelope box to perform collision detection between the various arm rods of the robot arm, the flexible grasping unit and the stereo camera are described by an axisymmetric envelope box to perform collision detection between the tool and the camera, and the obstacle is described by an axisymmetric envelope box to perform collision detection between the robot arm and the obstacle.

[0016] Preferably, calculating the path from the starting position to the target position includes:

[0017] S1, for the robot arm, the stereo camera and the flexible grasping unit described by the envelope box, the starting position of the path planning is used as the initial point X init , from the initial point X init Start to generate a random tree T, where the initial point X init is the root node of the random tree;

[0018] S2, generate random sampling pose points X in the collision-free area that has not been visited with a predetermined probability P rand , used to guide subsequent random tree expansion;

[0019] S3, calculate the existing nodes on the traversal random tree T and the randomly sampled pose points X rand The Euclidean distance of the randomly sampled pose point X is determined rand The nearest node X nearest ;

[0020] S4, from node X nearest Towards a randomly sampled pose point X rand The direction is expanded with a predetermined step size to generate a new node X new If an obstacle is encountered during the expansion process, the current expansion is stopped and the process returns to S2 until it is connected to the target position and the path from the starting position to the target position is obtained.

[0021] The present invention also provides an autonomous obstacle avoidance grasping method, wherein the method comprises:

[0022] Using a control unit to describe an envelope box for a robotic arm, a stereo camera, a flexible grasping unit and an obstacle to perform self-collision detection and obstacle collision detection, wherein the robotic arm is arranged on a table, the stereo camera is arranged on the robotic arm, and the flexible grasping unit is arranged at the end of the robotic arm;

[0023] Using the stereo camera to obtain point cloud images and plane images of the target object and obstacles;

[0024] Using the control unit to obtain point cloud data of the target object and the obstacle according to the point cloud image and the plane image of the target object and the obstacle;

[0025] The control unit is used to process the point cloud data of the target object and the obstacle to obtain the position of the grasping point;

[0026] In the absence of collision, a control unit is used to calculate a path from a starting position to a target position;

[0027] The control unit is used to control the mechanical arm to grasp the target object according to the calculated path.

[0028] Through the above technical solution, the flexible unit can be used to adapt to regular and irregular target objects, and by constructing obstacle entities and target grasping positions, the obstacle avoidance path can be planned, and the robot arm can be guided to safely bypass obstacles and grasp targets. In this way, the positioning and grasping of different target objects can be achieved. During the recognition process, obstacles and targets can be identified to plan obstacle avoidance grasping paths, and collisions between the robot arm, tools and obstacles and collisions between the robot arm and itself can be prevented. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The included drawings are used to provide a further understanding of the embodiments of the present invention, which constitute a part of the specification, are used to illustrate the embodiments of the present invention, and together with the text description, explain the principles of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 It shows the overall composition diagram of an autonomous obstacle avoidance and grasping system according to an embodiment of the present invention;

[0031] Figure 2 An overall flow chart of an autonomous obstacle avoidance and grasping method according to an embodiment of the present invention is shown;

[0032] Figure 3 A flow chart of an obstacle avoidance path planning method according to an embodiment of the present invention is shown;

[0033] Figure 4 A target grasping position calculation flow chart according to an embodiment of the present invention is shown;

[0034] Figure 5 A simulation diagram of the obstacle avoidance and grasping method according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0035] It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0037] Unless otherwise specifically stated, the relative arrangement of the parts and steps described in these embodiments, numerical expressions and numerical values ​​do not limit the scope of the present invention. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be regarded as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0038] Figure 1 The figure shows the overall composition of an autonomous obstacle avoidance and grasping system according to an embodiment of the present invention.

[0039] like Figure 1As shown, an embodiment of the present invention provides an autonomous obstacle avoidance grasping system, wherein the system comprises a mechanical arm 1, a stereo camera 2, a flexible grasping unit 3 and a control unit, wherein the mechanical arm 1 is arranged on a table 4, the stereo camera 2 is arranged on the mechanical arm 1, and the flexible grasping unit 3 is arranged at the end of the mechanical arm 1.

[0040] The stereo camera 2 is used to obtain point cloud images and plane images of the target object 6 and the obstacle 5;

[0041] The control unit is used to obtain point cloud data of the target object 6 and the obstacle 5 based on the point cloud image and the plane image of the target object 6 and the obstacle 5, process the point cloud data of the target object 6 and the obstacle 5 to obtain the position of the grasping point, perform an envelope box description on the robotic arm 1, the stereo camera 2, the flexible grasping unit 3 and the obstacle 5 to perform self-collision detection and obstacle collision detection, and calculate the path from the starting position to the target position when there is no collision, and control the robotic arm 1 to grasp the target object 6 according to the calculated path.

[0042] The robot arm may be a six-axis robot arm, and the flexible gripping unit may be a flexible gripping clamp. For example, the flexible gripping clamp may be adapted to grip regular or irregular objects within a range of 10 cm. Figure 1 For example, the obstacle may be a cubic obstacle and the target may be a sphere.

[0043] Through the above technical solution, the flexible unit can be used to adapt to regular and irregular target objects, and by constructing obstacle entities and target grasping positions, the obstacle avoidance path can be planned, and the robot arm can be guided to safely bypass obstacles and grasp targets. In this way, the positioning and grasping of different target objects can be achieved. During the recognition process, obstacles and targets can be identified to plan obstacle avoidance grasping paths, and collisions between the robot arm, tools and obstacles and collisions between the robot arm and itself can be prevented.

[0044] According to an embodiment of the present invention, obtaining point cloud data of the target object 6 and the obstacle 5 according to the point cloud image and the plane image of the target object 6 and the obstacle 5 includes:

[0045] Conditional filtering is performed on the point cloud image and plane image of the target object 6 and the obstacle 5, plane segmentation is performed on the filtered image to remove the platform surface, and Euclidean distance clustering segmentation is performed on the image after the platform surface is removed to obtain point cloud data of the target object 6 and the obstacle 5.

[0046] According to an embodiment of the present invention, obtaining point cloud data of the target object 6 and the obstacle 5 according to the point cloud image and the plane image of the target object 6 and the obstacle 5 includes:

[0047] Based on deep learning target detection, the target object 6 and the obstacle 5 are located according to their point cloud images and plane images, and the areas of the target object 6 and the obstacle 5 are segmented to obtain the point cloud data of the target object 6 and the obstacle 5.

[0048] According to an embodiment of the present invention, processing the point cloud data of the target object 6 and the obstacle 5 to obtain the position of the grasping point includes:

[0049] The point cloud data of the target object 6 is fitted or registered, and the point cloud data of the obstacle 5 is enveloping, so as to obtain the position of the grasping point.

[0050] According to one embodiment of the present invention, self-collision detection includes collision detection between the arms of the robotic arm 1 and collision detection between the flexible grasping unit 3 and the stereo camera 2; obstacle collision detection includes collision detection between the robotic arm 1 and the obstacle 5.

[0051] According to an embodiment of the present invention, performing an envelope box description on the robot arm 1, the stereo camera 2, the flexible grasping unit 3 and the obstacle 5 to perform self-collision detection and obstacle collision detection includes:

[0052] The robot arm is described by a cylindrical envelope box to perform collision detection between the arms of the robot 1, the flexible grasping unit 3 and the stereo camera 2 are described by an axisymmetric envelope box to perform collision detection between the tool and the camera, and the obstacle 5 is described by an axisymmetric envelope box to perform collision detection between the robot 1 and the obstacle 5.

[0053] According to an embodiment of the present invention, calculating a path from a starting position to a target position includes:

[0054] S1, for the robot arm 1, the stereo camera 2 and the flexible gripping unit 3 described by the envelope box, the starting position of the path planning is used as the initial point X init , from the initial point X init Start to generate a random tree T, where the initial point X init is the root node of the random tree;

[0055] S2, generate random sampling pose points X in the collision-free area that has not been visited with a predetermined probability P rand , used to guide subsequent random tree expansion;

[0056] S3, calculate the existing nodes on the traversal random tree T and the randomly sampled pose points X rand The Euclidean distance of the randomly sampled pose point X is determined rand The nearest node X nearest ;

[0057] S4, from node X nearest Towards a randomly sampled pose point X ran The direction of d is expanded with a predetermined step size to generate a new node X new If an obstacle is encountered during the expansion process, the current expansion is stopped and the process returns to S2 until it is connected to the target position and the path from the starting position to the target position is obtained.

[0058] like Figure 2 As shown, an embodiment of the present invention provides an autonomous obstacle avoidance grasping method, wherein the method includes:

[0059] Using a control unit to describe an envelope box for a robotic arm, a stereo camera, a flexible grasping unit and an obstacle to perform self-collision detection and obstacle collision detection, wherein the robotic arm is arranged on a table, the stereo camera is arranged on the robotic arm, and the flexible grasping unit is arranged at the end of the robotic arm;

[0060] Using the stereo camera to obtain point cloud images and plane images of the target object and obstacles;

[0061] Using the control unit to obtain point cloud data of the target object and the obstacle according to the point cloud image and the plane image of the target object and the obstacle;

[0062] The control unit is used to process the point cloud data of the target object and the obstacle to obtain the position of the grasping point;

[0063] In the absence of collision, a control unit is used to calculate a path from a starting position to a target position;

[0064] The control unit is used to control the mechanical arm to grasp the target object according to the calculated path.

[0065] Through the above technical solution, the flexible unit can be used to adapt to regular and irregular target objects, and by constructing obstacle entities and target grasping positions, the obstacle avoidance path can be planned, and the robot arm can be guided to safely bypass obstacles and grasp targets. In this way, the positioning and grasping of different target objects can be achieved. During the recognition process, obstacles and targets can be identified to plan obstacle avoidance grasping paths, and collisions between the robot arm, tools and obstacles and collisions between the robot arm and itself can be prevented.

[0066] Specifically, Figure 2As shown, the robot arm, camera, gripper and obstacle can first be described by an envelope box for subsequent self-collision detection and obstacle collision detection. Preferably, a cylindrical envelope can be used for the robot arm to more accurately describe the posture of the robot arm, and an axisymmetric envelope method (AABB) envelope can be used for the camera, gripper and obstacle to reduce the amount of calculation. Obstacles and targets are photographed by a stereo camera. Preferably, the areas where obstacles and targets are located can be determined by a deep learning target detection method to quickly perform preliminary segmentation on the point cloud, and then the grasping position of the target is obtained by filtering, segmenting, fitting and registration.

[0067] A hierarchical envelope box detection algorithm is used for the robot's self-collision and collision with obstacles. For example, the commonly used regular spatial geometry can be used to envelop and abstract the robot and obstacles, so that the problem of collision detection between the robot and obstacles is transformed into the problem of judging whether regular geometric bodies in space overlap. The axisymmetric envelope method (AABB) is used for collision detection between the robot and obstacles. AABB uses a hexahedron to replace the object to be detected. The sides of the selected hexahedron are parallel to the world coordinate axis, which can completely accommodate the object to be detected and minimize the volume. The hexahedron enclosed by AABB can be expressed in space using a mathematical expression as follows:

[0068] R={(x,y,z)|l x ≤x≤u x ,l y ≤y≤u y ,l z ≤z≤u z}

[0069] In the formula, R represents the set of points of the hexahedron, (x, y, z) represents the projection coordinates of the points of the hexahedron on the x, y, and z axes, and l x ,u x ,l y ,u y ,l z ,u z They represent the minimum and maximum values ​​of the projection of the hexahedron on the x, y, and z axes respectively. For the collision detection of two AABBs, the projections on the three coordinate axes can be calculated respectively. If there is any direction where the projections do not overlap, then the two envelopes will not collide.

[0070] For self-collision detection, the cylinder envelope box collision detection method is used, and a cylinder that can completely envelop the detected object and has the smallest diameter is used to replace the object to be detected. The collision between two cylinders in space can be divided into three situations, namely, end-to-end collision, end-to-side collision, and side-to-side collision. Find the mother line of the cylinder, and make judgments based on the mother line in the order of center distance pre-detection → detection of end circle intersection → detection of side intersection. Eliminate the three collision situations respectively, and it can be determined that no collision occurs.

[0071] The above description of collision detection is merely exemplary and is not intended to limit the present invention. The present invention may also use other existing methods in the prior art to perform collision detection.

[0072] Based on the above collision detection judgment, such as Figure 3 It is the process of generating and processing the obstacle avoidance path. For example, the RRT algorithm can be used for obstacle avoidance path planning. The specific implementation process is as follows:

[0073] (1) Take the starting point of path planning as the initial point X init , generating a random tree T from the initial point, where the initial point X init is the root node of the random tree;

[0074] (2) Generate a random sampling pose point X in the collision-free area that has not been visited with a predetermined probability P ran d, used to guide the subsequent random tree expansion;

[0075] (3) Calculate the existing nodes and points X on the traversal random tree T ran The Euclidean distance of d, find the distance X ran d The nearest node X neares t;

[0076] (4) From X neares tTowards direction X ran d is expanded at a predetermined step size to generate a new node X new If an obstacle is encountered during the expansion process, the expansion is stopped and the process returns to step (2) and is repeated until the target point is connected.

[0077] The process of calculating the position of the target grasping point based on the point cloud image and plane image obtained by the stereo camera is as follows: Figure 4 As shown in the figure, after conditional filtering according to the range in the xyz coordinate axis direction, plane segmentation is performed to remove the platform surface, and then Euclidean distance is performed for cluster segmentation (also called distance segmentation), and finally AABB envelope is performed on the obstacle, and the segmented spherical target is fitted. Alternatively, obstacles and spheres can be quickly located based on deep learning target detection, and subsequent operations can be performed after the point cloud areas of obstacles and spheres are segmented.

[0078] like Figure 5 The test results of the autonomous obstacle avoidance and grasping system and method of the present invention are as follows: Figure 5 The robot arm obstacle avoidance method is simulated using simulation software. The obstacle avoidance method can better reach the target grasping position by gradually iterating by generating random path points. In the actual obstacle avoidance and grasping process, the robot arm can better avoid obstacles and grasp the target sphere.

[0079] It can be seen from the above embodiments that the present invention provides a system and method for autonomous obstacle avoidance and target object grasping by a robot arm based on vision guidance. The system uses a stereo camera to shoot obstacle and target point clouds and plane images. The flexible gripper at the end of the six-axis robot arm can grasp regular objects and special-shaped objects. The system and method can identify obstacles and targets, automatically plan the path from the starting point to bypass the obstacle and reach the target grasping position, thereby guiding the robot arm to grasp. More specifically, the system and method obtains point cloud data of the target object and the obstacle through a stereo camera, determines the position of the grasping point by filtering, segmenting, fitting or registering the point cloud of the target object, performs self-collision detection of the robot arm by describing the robot arm rod through a cylindrical bounding box, performs self-collision detection of the tool and the camera by describing the end gripper and the camera through an AABB bounding box, and performs collision detection between the robot arm and the obstacle by computing the axis-aligned bounding box (AABB) of the obstacle point cloud to describe the spatial position of the obstacle, and performs collision detection between the robot arm and the obstacle. Through obstacle collision detection and self-collision detection, the robot arm and the attached gripper, camera and other tools can be effectively avoided from colliding, thereby improving the safety of the equipment.

[0080] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, perpendicular, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the devices or elements referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention. The directional words "inside and outside" refer to the inside and outside relative to the contours of each component itself.

[0081] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used here to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0082] In addition, it should be noted that the use of terms such as "first" and "second" to limit components is only for the convenience of distinguishing the corresponding components. If not otherwise stated, the above terms have no special meaning and therefore cannot be understood as limiting the scope of protection of the present invention.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An autonomous obstacle avoidance grasping system, characterized in that: The system comprises a mechanical arm (1), a stereo camera (2), a flexible grasping unit (3) and a control unit, wherein the mechanical arm (1) is arranged on a table (4), the stereo camera (2) is arranged on the mechanical arm (1), and the flexible grasping unit (3) is arranged at the end of the mechanical arm (1). The stereo camera (2) is used to obtain point cloud images and plane images of the target object (6) and the obstacle (5); The control unit is used to obtain point cloud data of the target object (6) and the obstacle (5) based on the point cloud image and the plane image of the target object (6) and the obstacle (5), process the point cloud data of the target object (6) and the obstacle (5) to obtain the position of the grasping point, perform an envelope box description on the mechanical arm (1), the stereo camera (2), the flexible grasping unit (3) and the obstacle (5) to perform self-collision detection and obstacle collision detection, and calculate a path from a starting position to a target position in the absence of a collision, and control the mechanical arm (1) to grasp the target object (6) according to the calculated path.

2. The system according to claim 1, characterized in that Point cloud data of the target object (6) and the obstacle (5) are obtained according to the point cloud image and the plane image of the target object (6) and the obstacle (5), including: Conditional filtering is performed on the point cloud image and plane image of the target object (6) and the obstacle (5), plane segmentation is performed on the filtered image to remove the platform surface, and Euclidean distance clustering segmentation is performed on the image after the platform surface is removed to obtain point cloud data of the target object (6) and the obstacle (5).

3. The system according to claim 1, characterized in that Point cloud data of the target object (6) and the obstacle (5) are obtained according to the point cloud image and the plane image of the target object (6) and the obstacle (5), including: Based on deep learning target detection, the target object (6) and the obstacle (5) are located according to the point cloud image and the plane image of the target object (6) and the obstacle (5), and the areas of the target object (6) and the obstacle (5) are segmented to obtain the point cloud data of the target object (6) and the obstacle (5).

4. The system according to claim 2 or 3, characterized in that: The point cloud data of the target object (6) and the obstacle (5) are processed to obtain the positions of the grasping points including: The point cloud data of the target object (6) is fitted or registered, and the point cloud data of the obstacle (5) is enclosed by an obstacle envelope to obtain the position of the grasping point.

5. The system according to claim 1, characterized in that The self-collision detection includes the collision detection between the arms of the robot arm (1) and the collision detection between the flexible grasping unit (3) and the stereo camera (2); the obstacle collision detection includes the collision detection between the robot arm (1) and the obstacle (5).

6. The system according to claim 5, characterized in that Performing an envelope box description on the robot arm (1), the stereo camera (2), the flexible grasping unit (3) and the obstacle (5) to perform self-collision detection and obstacle collision detection includes: The robot arm rods are described by a cylindrical envelope box to perform collision detection between the various rods of the robot arm (1), the flexible grasping unit (3) and the stereo camera (2) are described by an axisymmetric envelope box to perform collision detection between the tool and the camera, and the obstacle (5) is described by an axisymmetric envelope box to perform collision detection between the robot arm (1) and the obstacle (5).

7. The system according to claim 6, characterized in that Calculating the path from the starting location to the target location includes: S1, for the robot arm (1), the stereo camera (2) and the flexible grasping unit (3) described by the envelope box, the starting position of the path planning is used as the initial point X init , from the initial point X init Start to generate a random tree T, where the initial point X init is the root node of the random tree; S2, generate random sampling pose points X in the collision-free area that has not been visited with a predetermined probability P rand , used to guide subsequent random tree expansion; S3, calculate the existing nodes on the traversal random tree T and the randomly sampled pose points X rand The Euclidean distance of the randomly sampled pose point X is determined rand The nearest node X nearest ; S4, from node X nearest Towards a randomly sampled pose point X rand The direction is expanded with a predetermined step size to generate a new node X new If an obstacle is encountered during the expansion process, the current expansion is stopped and the process returns to S2 until it is connected to the target position and the path from the starting position to the target position is obtained.

8. An autonomous obstacle avoidance grasping method, characterized in that: The method includes: Using a control unit to describe an envelope box for a robotic arm, a stereo camera, a flexible grasping unit and an obstacle to perform self-collision detection and obstacle collision detection, wherein the robotic arm is arranged on a table, the stereo camera is arranged on the robotic arm, and the flexible grasping unit is arranged at the end of the robotic arm; Using the stereo camera to obtain point cloud images and plane images of the target object and obstacles; Using the control unit to obtain point cloud data of the target object and the obstacle according to the point cloud image and the plane image of the target object and the obstacle; The control unit is used to process the point cloud data of the target object and the obstacle to obtain the position of the grasping point; In the absence of collision, a control unit is used to calculate a path from a starting position to a target position; The control unit is used to control the mechanical arm to grasp the target object according to the calculated path.

Images