公开(公告)号：US20220164653A1

公开(公告)日：2022-05-26

申请号：US17434855

申请日：2020-01-07

Applicant: SONY GROUP CORPORATION

Inventor： YUKA ARIKI ,  TAKUYA NARIHIRI

IPC: G06N3/08 ,  G06F16/903

Abstract: Provided is an information processing apparatus including a learning unit that learns a heuristic function related to a path search by using a convolutional neural network. The learning unit includes a first convolutional neural network that outputs a first feature value based on an environment map, and a second convolutional neural network that outputs a second feature value related to an internal state of a search subject, performs learning using a loss related to a concatenated value of the first feature value and the second feature value, and outputs a heuristic map in which the heuristic function is represented as a two- or higher-dimensional image. The internal state of the search subject includes at least one element that has a degree of freedom different from a position in an environment.