Collision handling methods in grasp generation

Patent metadata

Field	Value
Publication number	US-12017356-B2
Application number	US-202117538380-A
Country	US
Kind code	B2
Filing date	Nov 30, 2021
Priority date	Nov 30, 2021
Publication date	Jun 25, 2024
Grant date	Jun 25, 2024

Abstract

A robotic grasp generation technique for part picking applications. Part and gripper geometry are provided as inputs, typically from CAD files. Gripper kinematics are also defined as an input. A set of candidate grasps is provided using any known preliminary grasp generation tool. A point model of the part and a model of the gripper contact surfaces with a clearance margin are used in an optimization computation applied to each of the candidate grasps, resulting in an adjusted grasp database. The adjusted grasps optimize grasp quality using a virtual gripper surface, which positions the actual gripper surface a small distance away from the part. A signed distance field calculation is then performed on each of the adjusted grasps, and those with any collision between the gripper and the part are discarded. The resulting grasp database includes high quality collision-free grasps for use in a robotic part pick-and-place operation.

First claim

What is claimed is:

1. A method for generating a collision-free grasp database for use by a robotic grasping system, said method comprising:
providing an object model including three-dimensional (3D) shape data for an object to be grasped, and a gripper model including 3D shape data and finger kinematics for a gripper;
providing a database of candidate grasps of the gripper on the object;
defining a gripper virtual contact surface which is offset from a gripper finger contact surface on each finger of the gripper based on a predefined clearance margin;
performing a grasp optimization on each of the candidate grasps to produce a database of adjusted grasps, using a computer having a processor and memory, where the grasp optimization maximizes a grasp quality value using the gripper virtual contact surfaces and a point cloud representation of the object model;
performing a signed distance field collision check on each of the adjusted grasps using the point cloud representation of the object model and a signed distance field grid of the gripper; and
providing the collision-free grasp database including each of the adjusted grasps which are determined to be collision-free in the signed distance field collision check.

2. The method according to claim 1 wherein the gripper virtual contact surface is offset from the gripper finger contact surface in a direction in which each gripper finger moves when grasping an object.

3. The method according to claim 1 further comprising using the collision-free grasp database during live robotic operations to identify a target object to grasp from a container of objects, including identifying the target object by mapping grasps from the collision-free grasp database onto estimated object poses from camera or sensor data, and providing target object grasp data to a robot controller which instructs a robot fitted with the gripper to grasp and move the target object.

4. The method according to claim 1 wherein each candidate grasp, adjusted grasp and collision-free grasp defines a pose of the gripper in a coordinate frame of the object.

5. The method according to claim 4 wherein the pose of the gripper in the coordinate frame of the object includes three positions and three rotations, and a gripper width.

6. The method according to claim 1 wherein the grasp optimization is formulated by defining sampling points on the gripper virtual contact surfaces and the point cloud representation of the object model, including a nearest neighbor point on the object corresponding with each of the sampling points on the gripper virtual contact surfaces.

7. The method according to claim 1 wherein the grasp optimization includes an objective function which converges to a locally-optimum grasp quality defined based on a total distance between each sampling point on the gripper virtual contact surfaces and its corresponding nearest neighbor point on the object, and includes a constraint equation which penalizes penetration of the gripper virtual contact surfaces into the object.

8. The method according to claim 7 wherein the grasp optimization using the gripper virtual contact surface results in an adjusted grasp in which the gripper finger contact surfaces are spaced from the object by the clearance margin.

9. The method according to claim 1 wherein performing the signed distance field collision check includes transforming the point cloud representation of the object model into a coordinate frame of the signed distance field grid of the gripper using an inverse of a pose defining the adjusted grasp.

10. The method according to claim 9 wherein performing the signed distance field collision check includes determining that the adjusted grasp is collision-free if no point in the point cloud representation of the object model falls within a grid cell of the signed distance field grid of the gripper.

11. The method according to claim 1 wherein the database of candidate grasps includes a large plurality of grasps, each defining a different position and orientation of the gripper with respect to the object, and which have not been evaluated for grasp quality or object-gripper interference.

12. A method for generating a collision-free grasp database, said method comprising:
providing a three-dimensional (3D) model of an object to be grasped, and a 3D model of a gripper;
providing a database of candidate grasps of the gripper on the object;
performing a grasp optimization on each of the candidate grasps to produce a database of adjusted grasps, using a computer having a processor and memory, where the grasp optimization maximizes a grasp quality value using a gripper virtual contact surface which is offset from a gripper finger contact surface on each finger of the gripper and a point cloud representation of the object model; and
performing a signed distance field collision check on each of the adjusted grasps to provide the collision-free grasp database including each of the adjusted grasps which are determined to be collision-free in the signed distance field collision check.

13. A robotic grasp generation system comprising:
a computer having a processor and memory, said computer configured to generate
a collision-free grasp database, including;
providing an object model including three-dimensional (3D) shape data for an object to be grasped, and a gripper model including 3D shape data and finger kinematics for a gripper;
providing a database of candidate grasps of the gripper on the object;
defining a gripper virtual contact surface which is offset from a gripper finger contact surface on each finger of the gripper based on a predefined clearance margin;
performing a grasp optimization on each of the candidate grasps to produce a database of adjusted grasps, where the grasp optimization maximizes a grasp quality value using the gripper virtual contact surfaces and a point cloud representation of the object model;
performing a signed distance field collision check on each of the adjusted grasps using the point cloud representation of the object model and a signed distance field grid of the gripper; and
providing the collision-free grasp database including each of the adjusted grasps which are determined to be collision-free in the signed distance field collision check.

14. The system according to claim 13 further comprising:
a 3D camera providing depth images to the computer of a container of objects during live robotic operations, where the computer identifies a target object to grasp from the container of objects by mapping grasps from the collision-free grasp database onto estimated object poses from the depth images;
a robot controller in communication with the computer and receiving grasp data about the target object; and
a robot fitted with the gripper to grasp and move the target object based on commands from the controller.

15. The system according to claim 13 wherein the gripper virtual contact surface is offset from the gripper finger contact surface in a direction in which each gripper finger moves when grasping an object.

16. The system according to claim 13 wherein each candidate grasp, adjusted grasp and collision-free grasp defines a pose of the gripper in a coordinate frame of the object, including three positions and three rotations, and a gripper width.

17. The system according to claim 13 wherein the grasp optimization is formulated by defining sampling points on the gripper virtual contact surfaces and the point cloud representation of the object model, including a nearest neighbor point on the object corresponding with each o

Assignees

• Fanuc Corp

Inventors

• Fan Yongxiang
• Lin Hsien-Chung

Classifications

• B25J9/1666

  Avoiding collision or forbidden zones · CPC title

• B25J9/1607

  Calculation of inertia, jacobian matrixes and inverses · CPC title

• B25J9/1653

  parameters identification, estimation, stiffness, accuracy, error analysis · CPC title

• B25J9/1697

  Vision controlled systems · CPC title

• B25J9/1669

  characterised by special application, e.g. multi-arm co-operation, assembly, grasping · CPC title