Pose determination by autonomous robots in a facility context
AssigneeGideon Brothers d.o.o.
InventorsJosip Cesic, Daniel Guja, Juraj Radic
Life of this patent
Estimating the term…
Abstract
A system and a method are disclosed where an autonomous robot captures an image of an object to be transported from a source to a destination. The robot generates a bounding box within the image surrounding the object. The robot applies a machine-learned model to the image with the bounding box, the machine-learned model configured to identify an object type of the object, and to identify features of the object based on the identified object type and the image. The robot determines which of the identified features of the object are visible to the autonomous robot, and determines a three-dimensional pose of the object based on the features determined to be visible to the autonomous robot.
Claims18
A non-transitory computer readable medium comprising memory with instructions encoded thereon that, when executed, cause one or more processors to perform operations, the instructions comprising instructions to:
- capture, by an autonomous robot, an image of an object to be transported from a source to a destination, wherein the image is a red-green-blue (RGB) image or a grayscale image;
- generate, by the autonomous robot, a bounding box within the image surrounding the object;
- reduce degrees of freedom of the object to four degrees of freedom by obscuring corner features of the object within the image until the object is constrained to a horizontal position;
- apply, by the autonomous robot, a machine-learned model to the image with the bounding box where the degrees of freedom of the object are reduced, the machine-learned model configured to identify an object type of the object, the object type referencing a category of the object out of a plurality of candidate categories, and to identify features of the object based on the identified object type and the image, at least some of the identified features being occluded and therefore not visible to the autonomous robot but nonetheless determinable by using the identified object type;
- determine, by the autonomous robot, which of the identified features of the object are visible to the autonomous robot;
- and
- determine, by the autonomous robot, a three-dimensional pose of the object based on the features determined to be visible to the autonomous robot.
A method comprising:
- capturing, by an autonomous robot, an image of an object to be transported from a source to a destination, wherein the image is a red-green-blue (RGB) image or a grayscale image;
- generating, by the autonomous robot, a bounding box within the image surrounding the object;
- reducing degrees of freedom of the object to four degrees of freedom by obscuring corner features of the object within the image until the object is constrained to a horizontal position;
- applying, by the autonomous robot, a machine-learned model to the image with the bounding box where the degrees of freedom of the object are reduced, the machine-learned model configured to identify an object type of the object, the object type referencing a category of the object out of a plurality of candidate categories, and to identify features of the object based on the identified object type and the image, at least some of the identified features being occluded and therefore not visible to the autonomous robot but nonetheless determinable by using the identified object type;
- determining, by the autonomous robot, which of the identified features of the object are visible to the autonomous robot;
- and
- determining, by the autonomous robot, a three-dimensional pose of the object based on the features determined to be visible to the autonomous robot.
A system comprising:
- memory with instructions encoded thereon;
- and
- one or more processors, that, when executing the instructions, are configured to perform operations comprising:
- capturing, by an autonomous robot, an image of an object to be transported from a source to a destination, wherein the image is a red-green-blue (RGB) image or a grayscale image;
- generating, by the autonomous robot, a bounding box within the image surrounding the object;
- reducing degrees of freedom of the object to four degrees of freedom by obscuring corner features of the object within the image until the object is constrained to a horizontal position;
- applying, by the autonomous robot, a machine-learned model to the image with the bounding box where the degrees of freedom of the object are reduced, the machine-learned model configured to identify an object type of the object, the object type referencing a category of the object out of a plurality of candidate categories, and to identify features of the object based on the identified object type and the image, at least some of the identified features being occluded and therefore not visible to the autonomous robot but nonetheless determinable by using the identified object type;
- determining, by the autonomous robot, which of the identified features of the object are visible to the autonomous robot;
- and
- determining, by the autonomous robot, a three-dimensional pose of the object based on the features determined to be visible to the autonomous robot.
Dependencies are inferred from the claim text.