多重代理人合作穩定性之研究

A Study on Stability Issues in Multiagent Cooperation

周祿敏(Lu-Min Chou)

指導教授 ： 劉立頌

國立中正大學/工學院/電機工程學系/碩士(2014年)

摘要

在智慧家庭中有許多的服務設備，其中使其自動化的領域已經研究已久。因此本研究不但要使其自動化，更要使其自主化，再提升服務設備運作的效率。提升設備運作效率的方向有很多，本研究選擇節能為研究方向，目標使服務設備在同環境、同狀態下，較其他設備節能。合作就是一個很好的方式，因此本研究使用代理人依附於服務設備上，去使設備具有思考的能力，以提升其運作的效率。再讓設備代理人間互相溝通、組成聯盟來進行合作，來使設備作進一步提升。本論文採用半競爭式合作的方式，使代理人在達到基礎條件的前提下，盡可能的去提升自己的利益。代理人在盡可能提升利益的情況下，很容易發生僅有些許微薄的利益，就轉而去和其他代理人合作，造成原聯盟的解散，導致代理人混亂並使系統負擔增加。其原因就是系統的穩定性不足，為此本系統採取賦稅法去降低代理人離開原本聯盟的可能性，以提升系統的穩定性。但賦稅法會隨著時間的經過變得不再適合系統當下的狀態。為此本研究加入個性參數，個性參數本身不僅僅可以提升系統穩定性，同時也可以彌補賦稅法受時間影響的問題。在本研究最後將系統實作於智慧家庭中以驗證研究的成果。

關鍵字

半競爭式合作 ； 代理人合作 ； 多重代理人 ； 穩定性 ； 賦稅法

並列摘要

Several service providers may coexist in a smart home, and many researchers have worked on how to let them become more autonomous. This thesis proposes a method to make the service providers work not only automatically, but also independently and efficiently. Efficiency may have many definitions, but we define it as the goal of all service providers to save energy while providing services. Cooperation is a proper way to reach this goal, so we use agents to control service providers, make the service providers work by themselves and communicate with each other, and form coalitions to work efficiently with other agents. In this study, we use semi-competitive agents which evaluate their own utility values to decide whether to work with others or not. Since each agent only cares for its own benefit, it joins a coalition that will result in a highest utility value. Every time an agent finds a coalition which offers more benefit, it leaves the original coalition for a new one even with a little gain. This will cause the system to be disordered and boost the system load unnecessarily since the stability of the system is not kept. To solve this situation, we use the taxation approach to reduce the willingness of the agents leaving their coalitions to another. However, the taxation approach will not suit the system after a period of time. We propose the concept of individual arguments which can not only increase the stability of the system, but also compensate the drawback of taxation. Finally, we implement the method on the system and use it in smart home.

並列關鍵字

semi-competitive ； agent cooperation ； multi-agent ； stability ； taxation

參考文獻

[2] Smarthome. [Online]. http://www.smarthome.org.nz/ [May. 10, 2014].

[3] eHome. [Online]. http://www.e-homeautomation.com/ [May. 10, 2014].

[8] J. Ferber, “Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence”, Addison-Wesly, 1999

[9] N. R. Jennings, “On agent-based software engineering,” Artificial intelligence, vol. 117, no. 2, pp. 277–296, 2000.

[11] N. R. Jennings , et al. “Automated negotiation: prospects, methods and challenges,” Group Decision and Negotiation, vol. 10, no. 2, pp. 199–215, 2001.