US 12,131,483 B2
Device and method for training a neural network for controlling a robot for an inserting task
Oren Spector, Modiin Maccabim Reut (IL); and Dotan Di Castro, Haifa (IL)
Assigned to ROBERT BOSCH GMBH, Stuttgart (DE)
Filed by Robert Bosch GmbH, Stuttgart (DE)
Filed on Mar. 14, 2022, as Appl. No. 17/654,723.
Claims priority of application No. 10 2021 109 334.1 (DE), filed on Apr. 14, 2021.
Prior Publication US 2022/0335622 A1, Oct. 20, 2022
Int. Cl. B25J 9/16 (2006.01); B25J 13/08 (2006.01); G06T 7/20 (2017.01); G06T 7/70 (2017.01); G06V 10/774 (2022.01)
CPC G06T 7/20 (2013.01) [B25J 9/161 (2013.01); B25J 9/1633 (2013.01); B25J 9/1687 (2013.01); B25J 9/1697 (2013.01); B25J 13/08 (2013.01); G06T 7/70 (2017.01); G06V 10/774 (2022.01); G05B 2219/32335 (2013.01); G05B 2219/39001 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A method for training a neural network to derive, from an image of a camera mounted on a robot, a movement vector to insert an object into an insertion, comprising the following steps:
for each position of a plurality of positions in which the object held by the robot touches a plane in which the insertion is located:
controlling the robot to move to the position,
taking a camera image by the camera, and
labelling the camera image with a movement vector between the position in the plane and the insertion in the plane; and
training the neural network using the labelled camera images.