本文選題：多Agent系統(tǒng) + 聯(lián)盟形成��； 參考：《云南大學(xué)》2013年博士論文

【摘要】：多Agent系統(tǒng)是分布式人工智能領(lǐng)域的兩個(gè)重要研究分支之一,其研究已經(jīng)為智能機(jī)器人系統(tǒng)、云計(jì)算、搜索引擎、交通控制、網(wǎng)絡(luò)通信等諸多應(yīng)用領(lǐng)域核心問題的解決開辟了新的思路,并推動(dòng)了相關(guān)應(yīng)用領(lǐng)域中實(shí)際復(fù)雜系統(tǒng)的形成與發(fā)展。在多Agent系統(tǒng)研究領(lǐng)域,多個(gè)Agent如何進(jìn)行有效合作和效用分配的問題是該領(lǐng)域的核心問題之一�；诤献鞑┺睦碚撨M(jìn)行多Agent系統(tǒng)的聯(lián)盟形成與效用分配問題的研究已經(jīng)取得了諸多成果,其基于合作博弈理論的研究方法也得到了普遍的認(rèn)可。然而,由于多Agent系統(tǒng)本身的復(fù)雜性和合作博弈理論到具體應(yīng)用的難點(diǎn),使得基于合作博弈理論對多Agent系統(tǒng)的聯(lián)盟形成和效用分配問題的研究仍然存在許多待解決的問題。針對現(xiàn)有研究仍待解決的問題,論文基于合作博弈理論對多Agent系統(tǒng)的聯(lián)盟形成與分配問題進(jìn)行了深入的研究,其研究工作具有一定的理論意義。論文具體的研究工作如下： (1)以公平分配為原則基于合作博弈理論中Shapley值的公平分配特性針對多Agent系統(tǒng)的動(dòng)態(tài)聯(lián)盟形成和分配問題進(jìn)行了研究。首先,論文提出了一種快速的靜態(tài)初始化聯(lián)盟形成算法。其次,針對Agent個(gè)體的參與能力和任務(wù)發(fā)生動(dòng)態(tài)變化的情況,論文提出了一種聯(lián)盟形成的快速動(dòng)態(tài)更新算法；針對Shapley值求解效率差的問題,論文提出一種快速求解聯(lián)盟中各個(gè)Agent近似Shapley值分配的算法。實(shí)驗(yàn)結(jié)果表明所提算法不僅獲得了預(yù)期聯(lián)盟形成結(jié)果和合理分配,而且具有較低的算法復(fù)雜度。 (2)以穩(wěn)定分配為原則基于合作博弈理論中談判集的穩(wěn)定分配特性對多Agent系統(tǒng)的聯(lián)盟形成與分配問題進(jìn)行了研究。針對求解談判集的復(fù)雜博弈問題,借助刪除談判劣勢聯(lián)盟后獲得的精簡聯(lián)盟集合,提出了一種基于遺傳算法的穩(wěn)定分配向量的求解算法,并給出了基于精簡聯(lián)盟集合獲得的談判集與經(jīng)典談判集相等的證明。論文進(jìn)一步提出了一種基于字典序比較尋找聯(lián)盟結(jié)構(gòu)集合中最穩(wěn)定聯(lián)盟結(jié)構(gòu)及穩(wěn)定分配向量的算法。實(shí)驗(yàn)結(jié)果表明所提出算法不僅保證了求解的成功率,而且具有較低的算法復(fù)雜度。 (3)針對多選擇合作博弈關(guān)于穩(wěn)定分配解理論的不足,論文將經(jīng)典合作博弈中談判集、內(nèi)核和核仁等與穩(wěn)定分配相關(guān)的概念拓展到多選擇合作博弈中,并證明了談判集的存在性、核仁的存在且唯一性、以及內(nèi)核、核仁和談判集三者之間的關(guān)系。借助拓展的穩(wěn)定分配解,論文提出了基于多選擇合作博弈求解復(fù)雜多Agent系統(tǒng)穩(wěn)定分配向量的方法,并給出了一種基于遺傳算法求解多個(gè)Agent在不同級別上穩(wěn)定分配向量的算法。
[Abstract]:Multiple Agent system is one of the two important research branches in the field of distributed artificial intelligence. It has been studied as intelligent robot system, cloud computing, search engine, traffic control, etc. The solution of many core problems in application fields such as network communication has opened up new ideas and promoted the formation and development of practical complex systems in related application fields. In the research field of multiple Agent systems, how to effectively cooperate and allocate the utility of multiple Agent is one of the core problems in this field. The research on alliance formation and utility allocation of multiple Agent systems based on cooperative game theory has made many achievements, and its research methods based on cooperative game theory have been generally accepted. However, due to the complexity of multiple Agent systems and the difficulties of cooperative game theory to specific application, there are still many problems to be solved in the research of alliance formation and utility allocation of multi-Agent systems based on cooperative game theory. Aiming at the problems that still remain to be solved, this paper makes a deep study on the formation and distribution of alliances in multiple Agent systems based on the cooperative game theory, which has a certain theoretical significance. The main work of this paper is as follows: 1) based on the principle of fair distribution, the paper studies the formation and allocation of dynamic alliances in multiple Agent systems based on the Shapley value in cooperative game theory. Firstly, a fast static initialization algorithm is proposed. Secondly, in view of the dynamic change of Agent individual's participation ability and task, this paper proposes a fast dynamic updating algorithm to form alliance, aiming at the problem of poor efficiency of solving Shapley value. In this paper, a fast algorithm to solve the Agent approximate Shapley value assignment in the alliance is proposed. Experimental results show that the proposed algorithm not only obtains the expected alliance formation results and reasonable allocation, Based on the stable assignment characteristics of negotiation sets in cooperative game theory, the formation and allocation of alliances in multi-Agent systems are studied. In order to solve the complex game problem of negotiation set, a stable assignment vector solving algorithm based on genetic algorithm is proposed by means of the reduced alliance set obtained after deleting the negotiating inferior alliance. The proof that the negotiation set based on the reduced alliance set is equal to the classical negotiation set is given. This paper further proposes an algorithm based on dictionary order comparison to find the most stable alliance structure and stable assignment vector in the set of alliance structures. The experimental results show that the proposed algorithm not only guarantees the success rate of the solution, but also has low computational complexity. The concepts of kernel and nucleolus which are related to stable distribution are extended to multi-choice cooperative game. The existence and uniqueness of negotiation set, the existence and uniqueness of nuclear kernel, and the relationship among kernel, nucleolus and negotiation set are proved. Based on the extended stable assignment solution, this paper proposes a method to solve the stable allocation vector of complex Agent system based on multi-selection cooperative game, and gives an algorithm based on genetic algorithm to solve the stable assignment vector of multiple Agent at different levels.
【學(xué)位授予單位】：云南大學(xué)
【學(xué)位級別】：博士
【學(xué)位授予年份】：2013
【分類號】：TP18

【參考文獻(xiàn)】

相關(guān)期刊論文 前4條

1 劉驚雷;童向榮;張偉;;一種快速構(gòu)建最優(yōu)聯(lián)盟結(jié)構(gòu)的方法[J];計(jì)算機(jī)工程與應(yīng)用;2006年04期

2 趙新宇;林作銓;;合同網(wǎng)協(xié)議中的Agent可信度模型[J];計(jì)算機(jī)科學(xué);2006年06期

3 羅翊，，石純一;Agent協(xié)作求解中形成聯(lián)盟的行為策略[J];計(jì)算機(jī)學(xué)報(bào);1997年11期

4 胡山立,石純一;一種任一時(shí)間聯(lián)盟結(jié)構(gòu)生成算法[J];軟件學(xué)報(bào);2001年05期

本文編號：2048741