US 12,093,051 B2
Path determination method
Yuji Hasegawa, Saitama (JP); and Sango Matsuzaki, Saitama (JP)
Assigned to HONDA MOTOR CO., LTD., Tokyo (JP)
Appl. No. 17/417,583
Filed by HONDA MOTOR CO., LTD., Tokyo (JP)
PCT Filed Aug. 7, 2019, PCT No. PCT/JP2019/031198
§ 371(c)(1), (2) Date Jun. 23, 2021,
PCT Pub. No. WO2020/136978, PCT Pub. Date Jul. 2, 2020.
Claims priority of application No. 2018-245255 (JP), filed on Dec. 27, 2018.
Prior Publication US 2022/0057804 A1, Feb. 24, 2022
Int. Cl. G05D 1/00 (2024.01)
CPC G05D 1/0221 (2013.01) [G05D 1/0214 (2013.01)] 2 Claims
OG exemplary drawing
 
1. A path determination method of determining a path in a case where an autonomous mobile robot moves to a destination under a condition in which a traffic participant including a pedestrian is present in a traffic environment to the destination, the path determination method comprising:
acquiring plural walking paths of a first pedestrian in a case where the first pedestrian walks toward a destination while avoiding interferences with plural second pedestrians other than the first pedestrian and where walking patterns of the plural second pedestrians are set to plural kinds of walking patterns which are different from each other;
creating plural databases of a relationship in which image data including an environment image representing a visual environment in front of the robot are associated with a behavior parameter representing behavior of the robot, the image data and the behavior parameter being obtained in a case where the robot moves along each of the plural walking paths;
creating a learned model as a behavior model subjected to learning by learning model parameters of the behavior model which has the image data as an input and has the behavior parameter as an output by a predetermined learning method by using the plural databases; and
determining the path of the robot by using the learned model,
wherein the plural databases are databases of a relationship in which the image data are associated with the behavior parameter, the image data and the behavior parameter being obtained in a case where the robot is assumed as a virtual robot existing in a virtual space and the plural walking paths are set in the virtual space, and the virtual robot moves along each of the plural walking paths in the virtual space.