WEKO3
アイテム
Modeling of Exception Management in Multi-Agent Systems
https://ir.soken.ac.jp/records/865
https://ir.soken.ac.jp/records/8659e5b08a1-2c06-4aee-b268-3aa123b9ac1d
名前 / ファイル | ライセンス | アクション |
---|---|---|
要旨・審査要旨 (280.7 kB)
|
||
本文 (15.4 MB)
|
Item type | 学位論文 / Thesis or Dissertation(1) | |||||
---|---|---|---|---|---|---|
公開日 | 2010-02-22 | |||||
タイトル | ||||||
タイトル | Modeling of Exception Management in Multi-Agent Systems | |||||
タイトル | ||||||
タイトル | Modeling of Exception Management in Multi-Agent Systems | |||||
言語 | en | |||||
言語 | ||||||
言語 | eng | |||||
資源タイプ | ||||||
資源タイプ識別子 | http://purl.org/coar/resource_type/c_46ec | |||||
資源タイプ | thesis | |||||
著者名 |
PLATON, Eric
× PLATON, Eric |
|||||
フリガナ |
プラトン, エリック
× プラトン, エリック |
|||||
著者 |
PLATON, Eric
× PLATON, Eric |
|||||
学位授与機関 | ||||||
学位授与機関名 | 総合研究大学院大学 | |||||
学位名 | ||||||
学位名 | 博士(情報学) | |||||
学位記番号 | ||||||
内容記述タイプ | Other | |||||
内容記述 | 総研大甲第1055号 | |||||
研究科 | ||||||
値 | 複合科学研究科 | |||||
専攻 | ||||||
値 | 17 情報学専攻 | |||||
学位授与年月日 | ||||||
学位授与年月日 | 2007-03-23 | |||||
学位授与年度 | ||||||
値 | 2006 | |||||
要旨 | ||||||
内容記述タイプ | Other | |||||
内容記述 | Multi-agent systems are often presented as the next mejor generation of software to cope with the increasing complexity in modern applications. MAS are ditstributed systems of autonomous and interacting entities named agents. They are possibly large-scale systems and the agent research community aims at having agents collaborate or compete with one another to achieve their functions in a highly modutar and flexible way. A variety of applications of agent technologies can be observed in state-of-the-art software developed from autonomous robots in manufacturing to software agents that assist users over the Internet. Multi-agent systems are therefore promising models and technologies in the future advances in Software engtneering and Artificial intelligence. <br /> Multi-agent systems are software tn the first place, and 50 years of history in Computer science has shown that constructing dependable systems requires dedicated endeavers and practices. Dependabillity refers to qualitties of a system, in terms of availability to the user of the system, reliabilty to provide the functions it is designed for, and safety and securtty of execution. Fault tolerance techniques were developed tn traditional Software engineering to increase the degree of dependabilty of software, and current achtevements allow guaranteeing several of the aforementioned qualities in many cases of close and homogeneous systems. Multi-agent systems challenge the current achtevements and target more complex systems, as required in the current demand from software users and the infrastructure of our society. Multi-agent systems target open and heterogeneous systems of autonomous agents. <br /> Among the techniques to increase the dependability of software systems, exception handling is notably famous for its strength and simplicity. Programming languages have for long exceptton handling capablities to process conveniently and systematicatty exceptional conditions encountered during a program execution. Distributed computing has however shown that exception handling required specific extensions in the case of distributed appli-cations, and work on software architectures and component-based development have shown the need for other modets as well Multi-agent systems set forth challenging properties that also need to reconsider the question of exception. <br /> The aim of thts thesis is to study the notion of exception in Multi-agent systems and to propose a framework adapted to the chllenges of openness, heterogeneity, and especially the autonomy of agents. Related work in the agent community has achieved in the past a number of results that showed the need for system-Level exception management tn Multi-agent systems. The management encompasses handling and the required mechanisms around the handling. The achievements to date set limitations on the type of MAS they can apply to. Agents are often not autonomous and the system-level approaches require agents to perfectly cllaborate in the exception management procedure. In this thesis, the ablitiy of agents to deal with exceptions by themselves in the first place is seen as a prerequisite to guarantee autonomy. Exception management then relies on agent-level mechanisms to cope with the shortcomings of current achievements and complement them. Agents keep the capability to freely choose when to initiate exception handling, and when to accept system-level support or rely on individual skills. <br /> The approach developed in this thesis ensures the autonomy of agents by a. novel execution model that guarantees the agent preserves control of itself all along its execution and despite the occurrence of exceptions. The model Lets the agent decide whether an event is an exception as an individual decision, thus enforcing further the autonomy. The model is formally described and a corresponding software architecture is proposed to implement it. The architecture is subsequently applied to a cese study to validate the approach, compare it to existing work, and evaluate its computational cost. The perspectives of this work lie in a number of challenges that can be further elaborated in the framework proposed in this thesis. In paticular, the automatic generation of handling strategies by agents in a range of situations is a promising capability that can expand the autonomy of agents in dealing with vartous exceptional situations. Another notable research direction is the evaluation by an agent of handling strategies received from other agents in the system. The interest in this topic is particularly relevant in future endeavors to bridge previous work, that essentially provide agents with strategies, with the present approach for autonomous agents that are abte to estimate when such an external support is acceptable. | |||||
所蔵 | ||||||
値 | 有 |