메뉴 건너뛰기
.. 내서재 .. 알림
소속 기관/학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색
질문

논문 기본 정보

자료유형
학술대회자료
저자정보
장우성 (연세대학교) 김종규 (연세대학교) 이수홍 (연세대학교)
저널정보
대한기계학회 대한기계학회 춘추학술대회 대한기계학회 2020년 학술대회
발행연도
2020.12
수록면
142 - 145 (4page)

이용수

표지
📌
연구주제
📖
연구배경
🔬
연구방법
🏆
연구결과
AI에게 요청하기
추천
검색
질문

초록· 키워드

오류제보하기
In this study, reinforcement learning is conducted on the swing and iron rod movement robots operated by the EV3 controller. There are two methods of performing. First, an operation in EV3 itself. Second, the local desktop acts as a client to exchange state and action values with each other through socket communication with a server built in EV3. On the other hand, for the two similar tasks, the swing and the iron bar, the swing robot applied relatively latest SAC (soft actor critic) algorithm that is applicable to the continuous action space. However, for iron rod robots, the problem was approached in a different way, viewed as a discrete action space problem, and applied Q-learning, an early reinforcement learning algorithm. At these two cases, the states, actions, and reward values for actions were defined in different ways. As a result, it was confirmed that the iron rod robot with Q learning converged in a short time and climbed high while riding the iron rod well. On the other hand, although the swing robot used the latest SAC algorithm, which is known to be relatively robust in learning convergence, the convergence property was not guaranteed within a short learning time. Accordingly, in this study, it was concluded that the latest algorithms do not always show good results, and that it is effective in reinforcement learning to appropriately reduce the dimensions of the problem into a discontinuous space depending on the situation.

목차

Abstract
1. 서론
2. 이론 및 실험
3. 결론
참고문헌

참고문헌 (0)

참고문헌 신청

함께 읽어보면 좋을 논문

논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!

이 논문의 저자 정보

이 논문과 함께 이용한 논문

최근 본 자료

전체보기

댓글(0)

0