멀티에이전트 기반 SCM 모델링 및 구현

김태운,양성민,서대희 한국정보시스템학회 2003 情報시스템硏究 Vol.12 No.2

The purpose of this paper is to propose the modeling of multiagent based SCM and implement the prototype in the Internet environment. SCM process follows the supply chain operations reference (SCOR) model which has been suggested by Supply Chain Counsil. SCOR model has been positioned to become the industry standard for describing and improving operational process in SCM. Five basic processes, plan, source, make, deliver and return are defined in the SCOR model, through which a company establishes its supply chain competitive objectives. A supply chain is a world wide network of suppliers, factories, warehouses, distribution centers and retailers through which raw materials are acquired, transformed or manufactured and delivered to customers by autonomous or semiautonomous process. With the pressure from the higher standard of customer compliance, a frequent model change, product complexity and globalization, the combination of supply chain process with an advanced infrastructure in terms of multiagent systems have been highly required. Since SCM is fundamentally concerned with coherence among multiple decision makers, a multiagent framework based on explicit communication between constituent agents such as suppliers, manufacturers, and distributors is a natural choice. Multiagent framework is defined to perform different activities within a supply chain. Dynamic and changing functions of supply chain can be dealt with multi-agent by cooperating with other agents. In the areas of inventory management, remote diagnostics, communications with field workers, order fulfillment including tracking and monitoring, stock visibility, real-time shop floor data collection, asset tracking and warehousing, customer-centric supply chain can be applied and implemented utilizing multiagent. In this paper, for the order processing event between the buyer and seller relationship, multiagent were defined corresponding to the SCOR process. A prototype system was developed and implemented on the actual TCP/IP environment for the purchase order processing event. The implementation result assures that multiagent based SCM enhances the speed, visibility, proactiveness and responsiveness of activities in the supply chain.

적대적 멀티 에이전트 환경에서 효율적인 강화 학습을 위한 정책 모델링

권기덕(Kiduk Kwon),김인철(Incheol Kim) 한국정보과학회 2008 정보과학회논문지 : 소프트웨어 및 응용 Vol.35 No.3

멀티 에이전트 강화 학습에서 해결해야 할 중요한 문제는 자신의 작업 성능에 영향을 미칠 수 있는 다른 에이전트들이 존재하는 동적 환경에서 한 에이전트가 시행착오적 상호작용을 통해 어떻게 자신의 최적 행동 정책을 학습할 수 있느냐 하는 것이다. 멀티 에이전트 강화 학습을 위한 기존 연구들은 대부분 단일 에이전트 MDP 기반의 강화 학습기법들을 큰 변화 없이 그대로 적용하거나 비록 다른 에이전트에 관한 별도의 모델을 이용하더라도 다른 에이전트에 관해 요구되는 정보나 가정이 현실적이지 못하다는 한계점을 가지고 있다. 본 논문에서는 멀티 에이전트 강화 학습기술에 기초가 되는 기본 개념들을 정형화하고 이들을 기초로 기존 연구들의 특징과 한계점을 비교한다. 그리고 새로운 행동 정책 모델을 소개한 뒤, 이것을 이용한 강화 학습 방법을 설명한다. 본 논문에서 제안하는 멀티 에이전트 강화학습 방법은 상대 모델을 이용하는 기존의 멀티 에이전트 강화 학습 연구들에서 주로 시도되었던 상대 에이전트의 Q 평가 함수 모델 대신 상대 에이전트의 행동 정책 모델을 학습하며, 표현력은 풍부하나 학습에 시간과 노력이 많이 요구되는 유한 상태 오토마타나 마코프 체인과 같은 행동 정책 모델들에 비해 비교적 간단한 형태의 행동정책 모델을 이용함으로써 학습의 효율성을 높였다. 또한, 본 논문에서는 대표적인 적대적 멀티 에이전트 환경인 고양이와 쥐게임을 소개하고, 이 게임을 테스베드삼아 비교 실험들을 수행하고 그 결과를 설명함으로써 본 논문에서 제안하는 정책 모델 기반의 멀티 에이전트 강화 학습의 효과를 분석해본다. An important issue in multiagent reinforcement learning is how an agent should learn its optimal policy through trial-and-error interactions in a dynamic environment where there exist other agents able to influence its own performance. Most previous works for multiagent reinforcement learning tend to apply single-agent reinforcement learning techniques without any extensions or are based upon some unrealistic assumptions even though they build and use explicit models of other agents. In this paper, basic concepts that constitute the common foundation of multiagent reinforcement learning techniques are first formulated, and then, based on these concepts, previous works are compared in terms of characteristics and limitations. After that, a policy model of the opponent agent and a new multiagent reinforcement learning method using this model are introduced. Unlike previous works, the proposed multiagent reinforcement learning method utilize a policy model instead of the Q function model of the opponent agent. Moreover, this learning method can improve learning efficiency by using a simpler one than other richer but time-consuming policy models such as Finite State Machines(FSM) and Markov chains. In this paper, the Cat and Mouse game is introduced as an adversarial multiagent environment. And effectiveness of the proposed multiagent reinforcement learning method is analyzed through experiments using this game as testbed.

효율적인 멀티 에이전트 강화 학습을 위한 나이브 베이지안 기반 상대 정책 모델

권기덕 ( Ki-duk Kwon ) 한국인터넷정보학회 2008 인터넷정보학회논문지 Vol.9 No.6

멀티 에이전트 강화학습에서 중요한 이슈 중의 하나는 자신의 성능에 영향을 미칠 수 있는 다른 에이전트들이 존재하는 동적 환경에서 어떻게 최적의 행동 정책을 학습하느냐 하는 것이다. 멀티 에이전트 강화 학습을 위한 기존 연구들은 대부분 단일 에이전트 강화 학습기법들을 큰 변화 없이 그대로 적용하거나 비록 다른 에이전트에 관한 별도의 모델을 이용하더라도 현실적이지 못한 가정들을 요구한다. 본 논문에서는 상대 에이전트에 대한 나이브 베이지안 기반의 행동 정책 모델을 소개한 뒤, 이것을 이용한 강화 학습 방법을 설명한다. 본 논문에서 제안하는 멀티 에이전트 강화학습 방법은 기존의 멀티 에이전트 강화 학습 연구들과는 달리 상대 에이전트의 Q 평가 함수 모델이 아니라 나이브 베이지안 기반의 행동 정책 모델을 학습한다. 또한, 표현력은 풍부하나 학습에 시간과 노력이 많이 요구되는 유한 상태 오토마타나 마코프 체인과 같은 행동 정책 모델들에 비해 비교적 간단한 형태의 행동 정책 모델을 이용함으로써 학습의 효율성을 높였다. 본 논문에서는 대표적인 적대적 멀티 에이전트 환경인 고양이와 쥐게임을 소개한 뒤, 이 게임을 테스트 베드 삼아 실험들을 전개함으로써 제안하는 나이브 베이지안 기반의 정책 모델의 효과를 분석해본다. An important issue in Multiagent reinforcement learning is how an agent should learn its optimal policy in a dynamic environment where there exist other agents able to influence its own performance. Most previous works for Multiagent reinforcement learning tend to apply single-agent reinforcement learning techniques without any extensions or require some unrealistic assumptions even though they use explicit models of other agents. In this paper, a Naive Bayesian based policy model of the opponent agent is introduced and then the Multiagent reinforcement learning method using this model is explained. Unlike previous works, the proposed Multiagent reinforcement learning method utilizes the Naive Bayesian based policy model, not the Q function model of the opponent agent. Moreover, this learning method can improve learning efficiency by using a simpler one than other richer but time-consuming policy models such as Finite State Machines(FSM) and Markov chains. In this paper, the Cat and Mouse game is introduced as an adversarial Multiagent environment. And then effectiveness of the proposed Naive Bayesian based policy model is analyzed through experiments using this game as test-bed.

A Multiagent System for Microgrid Operation in the Grid-interconnected Mode

Kim, Hak-Man,Kinoshita, Tetsuo The Korean Institute of Electrical Engineers 2010 Journal of Electrical Engineering & Technology Vol.5 No.2

This paper presents a multiagent system for microgrid operation in the grid-interconnected mode. An energy market environment with generation competition is adopted for microgrid operation in order to guarantee autonomous participation and meet the requirements of participants in the microgrid. The modified Contract Net Protocol (CNP) is used as a protocol for interactions among agents. The multiagent system for microgrid operation based on the modified CNP and the energy market environment is designed and implemented. To verify the feasibility of the suggested multiagent system, experiments on three operation conditions are carried out.

A New Modified CNP for Autonomous Microgrid Operation Based on Multiagent System

Hak-Man Kim,Wenpeng Wei,Tetsuo Kinoshita 대한전기학회 2011 Journal of Electrical Engineering & Technology Vol.6 No.1

This paper presents a new modified Contract Net Protocol (CNP) for microgrid operation based on multiagent system. The CNP is a widely used protocol for interactions among distributed problem solving. The Contract Net Interaction Protocol of the Foundation for Intelligent Physical Agents (FIPA-CNIP) is a minor modification of the original CNP for multiagent system applications. In this paper, a modified CNP (MCNP) based on the FIPA-CNIP is proposed for more specialized interactions among agents for microgrid operation. A multiagent system is designed and constructed for microgrid operation. A microgrid operation based on the multiagent system is tested to check the functionality of the proposed MCNP.

A New Modified CNP for Autonomous Microgrid Operation Based on Multiagent System

Kim, Hak-Man,Wei, Wenpeng,Kinoshita, Tetsuo The Korean Institute of Electrical Engineers 2011 Journal of Electrical Engineering & Technology Vol.6 No.1

This paper presents a new modified Contract Net Protocol (CNP) for microgrid operation based on multiagent system. The CNP is a widely used protocol for interactions among distributed problem solving. The Contract Net Interaction Protocol of the Foundation for Intelligent Physical Agents (FIPA-CNIP) is a minor modification of the original CNP for multiagent system applications. In this paper, a modified CNP (MCNP) based on the FIPA-CNIP is proposed for more specialized interactions among agents for microgrid operation. A multiagent system is designed and constructed for microgrid operation. A microgrid operation based on the multiagent system is tested to check the functionality of the proposed MCNP.

Talmudic Approach to Load Shedding of Islanded Microgrid Operation Based on Multiagent System

Hak-Man Kim,Tetsuo Kinoshita,Yujin Lim 대한전기학회 2011 Journal of Electrical Engineering & Technology Vol.6 No.2

This paper presents a load-shedding scheme using the Talmud rule in islanded microgrid operation based on a multiagent system. Load shedding is an intentional load reduction to meet a power balance between supply and demand when supply shortages occur. The Talmud rule originating from the Talmud literature has been used in bankruptcy problems of finance, economics, and communications. This paper approaches the load-shedding problem as a bankruptcy problem. A load-shedding scheme is mathematically expressed based on the Talmud rule. For experiment of this approach, a multiagent system is constructed to operate test islanded microgrids autonomously. The suggested loadshedding scheme is tested on the test islanded microgrids based on the multiagent system. Results of the tests are discussed.

A Multiagent System for Microgrid Operation in the Grid-interconnected Mode

Hak-Man Kim,Tetsuo Kinoshita 대한전기학회 2010 Journal of Electrical Engineering & Technology Vol.5 No.2

This paper presents a multiagent system for microgrid operation in the grid-interconnected mode. An energy market environment with generation competition is adopted for microgrid operation in order to guarantee autonomous participation and meet the requirements of participants in the microgrid. The modified Contract Net Protocol (CNP) is used as a protocol for interactions among agents. The multiagent system for microgrid operation based on the modified CNP and the energy market environment is designed and implemented. To verify the feasibility of the suggested multiagent system, experiments on three operation conditions are carried out.

문단 기반 멀티에이전트 메타 검색 엔진 설계

강남오,박사준,황수철 한국지식정보기술학회 2012 한국지식정보기술학회 논문지 Vol.7 No.2

As the number of documents published on the Internet increases, users are gradually difficult to retrieve information accurately and efficiently. In general, users visit web pages by using the result list generated from a search engine and search each web page of the browsed list to discover information. And these processes are necessary for an additional time and work. In addition, for improving the retrieved result, increasing the number of search engines used in the retrieval process produces many different result lists, and it makes much more difficult for users to search information based on an integrated searching list. For solving these problems, we introduce using the multiagent and the localized information in the result documents. In this paper, we propose a Paragraph-based Multiagent metasearch engine which produces local-centric result list based on the calculated paragraph-unit similarity in web documents, which is browsed by numerous search engines for a user query.

모바일 컴퓨팅 환경에서 천식 환자 관리를 위한 데이터마이닝 기법 기반 멀티에이전트 시스템

김성림,권준희,유선국,홍천수 대한의료정보학회 2004 Healthcare Informatics Research Vol.10 No.3

Objective : In the mobile computing environments, doctors or patients with mobile devices can access services at any time, any place. It makes chronic diseases, such as asthma care successively and in long-term. This paper presents a new multiagent system using data mining technique for the management of asthma in mobile computing environments. Methods : In mobile computing environments, mobile communication is prone to disconnection. To use the services when disconnected, we locally store the recommendation information in mobile devices. Then, when user reconnects network, the new information is delivered in mobile devices. Because mobile devices have a limitation of resource, we store only patterns and recent data. We adopted data mining technique from the large pool of accumulated patients data can be utilized to summarize observed correlations and trends. Results : We propose multiagent system composed of a Pattern Agent, a Selection Agent and a Recommendation Agent. They make asthmatic-caring system access effectively in mobile computing environments. We describe the underlying architecture and the process. Then we make a prototype with asthmatic patients data in the Severance Hospital in Korea. Conclusion : The results of our study can be contributed to aid management of asthma in mobile computing environments.