DISTRIBUTED LEARNING METHOD FOR NOT MULTI RUNNABLE SIMULATOR BASED INTELLIGENT AGENT

The present invention relates to a distributed learning method for intelligent agents using reinforcement learning. The method includes the steps of: applying a docker container to a distributed learning environment; and defining communication according to the application of the docker container. Ac...

Full description

Saved in:
Bibliographic Details
Main Authors CHOI HOJIN, LEE SUNG HU, KIM BORA, SIM HYUN WOO, NAM JEHYUN, YUN SUNG YEOL, CHOI HYUNG KYUN, LEE JUNG UK, LEE JONGWHOA, GU BON HONG, WON JUN HEE
Format Patent
LanguageEnglish
Korean
Published 06.12.2022
Subjects
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
ELECTRIC COMMUNICATION TECHNIQUE
ELECTRICITY
PHYSICS
TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHICCOMMUNICATION
Online AccessGet full text

Cover

More Information
Summary:The present invention relates to a distributed learning method for intelligent agents using reinforcement learning. The method includes the steps of: applying a docker container to a distributed learning environment; and defining communication according to the application of the docker container. According to the present invention, it is possible to quickly accumulate experience data even in a simulator environment in which multiple executions are impossible without greatly modifying an existing algorithm. 본 발명은 강화 학습을 이용하여 지능형 에이전트를 학습하는 분산 학습 방법에 있어서, 도커 컨테이너를 분산 학습 환경에 적용하는 적용 단계 및 상기 도커 컨테이너의 적용에 따른 통신을 정의하는 정의 단계를 포함한다. 본 발명에 의하면 기존의 알고리즘을 크게 변형하지 않으면서 다중 실행이 불가능한 시뮬레이터 환경에서도 빠르게 경험 데이터를 축적할 수 있게 된다.
Bibliography:Application Number: KR20210068564