DIGITAL LIBRARY ARCHIVE
HOME > DIGITAL LIBRARY ARCHIVE
< Previous   List   Next >  
L-CAA : An Architecture for Behavior-Based Reinforcement Learning
Full-text Download
Jonggeun Hwang (Dept. of Computer Science, Kyonggi University)
InCheol Kim (Dept. of Computer Science, Kyonggi University)
Vol. 14, No. 3, Page: 59 ~ 76
Keywords
Agent Architecture, Reinforcement Learning, Q Function Value, UTBot
Abstract
In this paper, we propose an agent architecture called L-CAA that is quite effective in real-time dynamic environments. L-CAA is an extension of CAA, the behavior-based agent architecture which was also developed by our research group. In order to improve adaptability to the changing environment, it is extended by adding reinforcement learning capability. To obtain stable performance, however, behavior selection and execution in the L-CAA architecture do not entirely rely on learning. In L-CAA, learning is utilized merely as a complimentary means for behavior selection and execution. Behavior selection mechanism in this architecture consists of two phases. In the first phase, the behaviors are extracted from the behavior library by checking the user-defined applicable conditions and utility of each behavior. If multiple behaviors are extracted in the first phase, the single behavior is selected to execute in the help of reinforcement learning in the second phase. That is, the behavior with the highest expected reward is selected by comparing Q values of individual behaviors updated through reinforcement learning. L-CAA can monitor the maintainable conditions of the executing behavior and stop immediately the behavior when some of the conditions fail due to dynamic change of the environment. Additionally, L-CAA can suspend and then resume the current behavior whenever it encounters a higher utility behavior. In order to analyze effectiveness of the L-CAA architecture, we implement an L-CAA-enabled agent autonomously playing in an Unreal Tournament game that is a well-known dynamic virtual environment, and then conduct several experiments using it.
Show/Hide Detailed Information in Korean
L-CAA : 행위 기반 강화학습 에이전트 구조
황종근 (경기대학교 컴퓨터과학과)
김인철 (경기대학교 컴퓨터과학과)
Abstract
본 논문에서는 실시간 동적 환경에 효과적인 L-CAA 에이전트 구조를 제안한다. L-CAA 에이전트 구조는 변화하는 환경에 대한 적응성을 높이기 위해, 선행 연구를 통해 개발된 행위기반 에이전트 구조인 CAA에 강화학습 기능을 추가하여 확장한 것이다. 안정적인 성능을 위해 L-CAA 구조에서는 행위 선택과 실행을 학습에 전적으로 의존하지 않고 학습을 보조적으로 이용한다. L-CAA에서 행위 선택 메커니즘은 크게 두 단계로 나뉜다. 첫 번째 단계에서는 사용자가 미리 정의한 각 행위의 적용 가능 조건과 효용성을 검사함으로써 행위 라이브러리로부터 실행할 행위들을 추출한다. 하지만 첫 번째 단계에서 다수의 행위가 추출되면, 두 번째 단계에서는 강화학습의 도움을 받아 이들 중에서 실행 할 하나의 행위를 선택한다. 즉, 강화학습을 통해 갱신된 각 행위들의 Q 함수값을 서로 비교함으로써, 가장 큰 기대 보상값을 가진 행위를 선택하여 실행한다. 또한 L-CAA에서는 실행 중인 행위의 유지 가능 조건을 지속적으로 검사하여 환경의 동적 변화로 인해 일부 조건이 만족되지 않는 경우가 발생하면 현재 행위의 실행을 즉시 종료할 수 있다. 그 뿐 아니라, L-CAA는 행위 실행 중에도 효용성이 더 높은 다른 행위가 발생하면 현재의 행위를 일시 정지하였다가 복귀하는 기능도 제공한다. 본 논문에서는 L-CAA 구조의 효과를 분석하기 위해, 대표적인 동적 가상환경인 Unreal Tournament 게임에서 자율적으로 동작하는 L-CAA 기반의 에이전트를 구현하고, 이를 이용한 성능 실험을 전개해본다.
Cite this article
JIIS Style
Hwang, J., and I. Kim, "L-CAA : An Architecture for Behavior-Based Reinforcement Learning", Journal of Intelligence and Information Systems, Vol. 14, No. 3 (2008), 59~76.

IEEE Style
Jonggeun Hwang, and InCheol Kim, "L-CAA : An Architecture for Behavior-Based Reinforcement Learning", Journal of Intelligence and Information Systems, vol. 14, no. 3, pp. 59~76, 2008.

ACM Style
Hwang, J., and Kim, I., 2008. L-CAA : An Architecture for Behavior-Based Reinforcement Learning. Journal of Intelligence and Information Systems. 14, 3, 59--76.
Export Formats : BiBTeX, EndNote

Warning: include(/home/hosting_users/ev_jiisonline/www/admin/archive/advancedSearch.php) [function.include]: failed to open stream: No such file or directory in /home/hosting_users/ev_jiisonline/www/archive/detail.php on line 429

Warning: include() [function.include]: Failed opening '/home/hosting_users/ev_jiisonline/www/admin/archive/advancedSearch.php' for inclusion (include_path='.:/usr/local/php/lib/php') in /home/hosting_users/ev_jiisonline/www/archive/detail.php on line 429
@article{Hwang:JIIS:2008:337,
author = {Hwang, Jonggeun and Kim, InCheol},
title = {L-CAA : An Architecture for Behavior-Based Reinforcement Learning},
journal = {Journal of Intelligence and Information Systems},
issue_date = {September 2008},
volume = {14},
number = {3},
month = Sep,
year = {2008},
issn = {2288-4866},
pages = {59--76},
url = {},
doi = {},
publisher = {Korea Intelligent Information System Society},
address = {Seoul, Republic of Korea},
keywords = { Agent Architecture, Reinforcement Learning, Q Function Value and UTBot },
}
%0 Journal Article
%1 337
%A Jonggeun Hwang
%A InCheol Kim
%T L-CAA : An Architecture for Behavior-Based Reinforcement Learning
%J Journal of Intelligence and Information Systems
%@ 2288-4866
%V 14
%N 3
%P 59-76
%D 2008
%R
%I Korea Intelligent Information System Society