US 12,296,468 B2
Control device, inspection system, control method, and non-transitory computer-readable medium
Yasutaka Fuke, Tokyo (JP); Hirone Komatsu, Tokyo (JP); Takashi Matsuzawa, Tokyo (JP); Shin Asano, Tokyo (JP); Kotaro Tadano, Tokyo (JP); Koshiro Irie, Tokyo (JP); and Chanvicharo Ly, Tokyo (JP)
Assigned to MITSUBISHI HEAVY INDUSTRIES, LTD., Tokyo (JP); and TOKYO INSTITUTE OF TECHNOLOGY, Tokyo (JP)
Filed by MITSUBISHI HEAVY INDUSTRIES, LTD., Tokyo (JP); and TOKYO INSTITUTE OF TECHNOLOGY, Tokyo (JP)
Filed on Jan. 30, 2023, as Appl. No. 18/103,104.
Claims priority of application No. 2022-014816 (JP), filed on Feb. 2, 2022.
Prior Publication US 2023/0241786 A1, Aug. 3, 2023
Int. Cl. B25J 18/06 (2006.01); B25J 9/16 (2006.01); B25J 13/08 (2006.01); G05B 19/4155 (2006.01); F16L 101/30 (2006.01)
CPC B25J 18/06 (2013.01) [B25J 9/1607 (2013.01); B25J 13/089 (2013.01); G05B 19/4155 (2013.01); F16L 2101/30 (2013.01); G05B 2219/40269 (2013.01)] 5 Claims
OG exemplary drawing
 
1. A control device for an inspection device comprising a long flexible robot that is formed by connecting a plurality of units and bendable at each unit with a desired curvature, and a posture actuator configured to be able to adjust a posture of the units, the control device comprising:
a target posture specifying unit configured to specify a target posture including positions of representative points of the respective units traveling along a predetermined route;
an operation amount determination unit configured to determine an operation amount for bringing the unit to the target posture using a learning model with the target posture of the unit as an input and an operation amount of the posture actuator as an output; and
a learning unit configured to construct the learning model on the basis of the position of the representative point acquired from a position measurement sensor configured to measure a three-dimensional position of the representative point and the operation amount of the posture actuator,
wherein the learning unit
obtains a Jacobian matrix representing a sensitivity matrix for converting a speed of a tip of each unit into an operation speed of the posture actuator from a forward kinematics model with the operation amount of the posture actuator as an input and the position of the representative point as an output,
determines an operation amount for learning for bringing the unit closer to the target posture using the Jacobian matrix, and
calculates the speed of the tip of each unit from the Jacobian matrix so that an error between the posture of the unit after an operation with the operation amount for learning and the target posture becomes small, and learns the learning model on the basis of the position of the representative point of the unit after inching and the operation amount of the posture actuator when the inching is performed manually using an operation device.