Title:
OPERATION PREDICTION SYSTEM AND OPERATION PREDICTION METHOD
Document Type and Number:
WIPO Patent Application WO/2018/225862
Kind Code:
A1
Abstract:
An automatic operation system (5) is provided with a plurality of learned imitation models (61) and a model selection unit (35). The learned imitation models (61) are constructed by causing imitation models (61) corresponding to a plurality of groups, which are obtained by classifying work history data by means of an automatic classification algorithm of a classification system (51), to perform machine learning for the respective groups. The work history data include data indicating a surrounding environment and data indicating manipulation of an operator in the surrounding environment. The model selection unit (35) selects one imitation model (61) among the plurality of imitation models (61) on the basis of a result of given data being classified by the automatic classification algorithm of the classification system (51), the given data indicating a surrounding environment. The automatic operation system (5) inputs data indicating the surrounding environment to the imitation model (61) selected by the model selection unit (35) and causes the imitation model (61) to predict manipulation of an operator with respect to the surrounding environment.
Inventors:
HASUNUMA HITOSHI
ENOMOTO MASAYUKI
FUJIMORI JUN
ENOMOTO MASAYUKI
FUJIMORI JUN
Application Number:
PCT/JP2018/022087
Publication Date:
December 13, 2018
Filing Date:
June 08, 2018
Export Citation:
Assignee:
KAWASAKI HEAVY IND LTD (JP)
International Classes:
G06N99/00; B25J9/22
Foreign References:
| JP2017030137A | 2017-02-09 |
Attorney, Agent or Firm:
KATSURAGAWA, Naoki (JP)
Download PDF: