【摘要】：在互聯(lián)網(wǎng)技術(shù)和計(jì)算機(jī)技術(shù)迅猛發(fā)展的時(shí)代,Web服務(wù)及其組合技術(shù)在學(xué)術(shù)界和工業(yè)界都得到了迅猛發(fā)展,作為開放式的業(yè)務(wù)提供方式得到廣泛承認(rèn)。Web服務(wù)有著極好的平臺(tái)無關(guān)性、互通性和復(fù)用性,為了滿足用戶需求,服務(wù)推薦和服務(wù)組合的方法迅速出現(xiàn)。但是這些過程中服務(wù)都只能等待第三方調(diào)用后才能發(fā)揮作用,服務(wù)數(shù)量的增多,使得很多服務(wù)長期未被調(diào)用。軟件Agent有著主動(dòng)、自治和進(jìn)化等特性,可以獲取第三方的需求信息,向其推薦自身,在組成多Agent系統(tǒng)后利用協(xié)同技術(shù)可以實(shí)現(xiàn)更強(qiáng)的功能。但是它卻沒有web服務(wù)的那種平臺(tái)無關(guān)性、互通性和靈活性。為了讓服務(wù)本身具有智能性,得到智能的web服務(wù)和可以協(xié)同的服務(wù)組合算法,并驗(yàn)證服務(wù)的組合效率和組合成功率。將Agent與web服務(wù)結(jié)合,服務(wù)抓取網(wǎng)絡(luò)中已有的用戶需求信息,實(shí)時(shí)地、主動(dòng)地向服務(wù)請(qǐng)求者推薦自身。拋棄處理靜態(tài)請(qǐng)求的平臺(tái)Jack和Jadex,選用Jade平臺(tái)來處理動(dòng)態(tài)組合。Agent封裝后的服務(wù)采用基于偏好的離線部分可觀測馬爾科夫決策服務(wù)組合算法(PBOL_POMDP)和啟發(fā)式的在線部分可觀測馬爾科夫決策服務(wù)組合算法(HO_POMDP)能高效、獨(dú)立地完成服務(wù)組合的過程,并將結(jié)果返回服務(wù)調(diào)用者。將求解結(jié)果與經(jīng)典AEMS2算法和引文中OPUMO1算法比較,驗(yàn)證了算法性能�；诖�,本文做了如下工作： (1)引入Agent擴(kuò)展web服務(wù),將互聯(lián)網(wǎng)中web服務(wù)組合模型擴(kuò)展為多Agent動(dòng)態(tài)協(xié)同服務(wù)組合系統(tǒng),總結(jié)基于多Agent系統(tǒng)的動(dòng)態(tài)協(xié)同的兩層功能結(jié)構(gòu),給出執(zhí)行服務(wù)組合的動(dòng)態(tài)協(xié)同服務(wù)組合系統(tǒng)框架,對(duì)比了傳統(tǒng)的兩層框架和擴(kuò)展后的多Agent動(dòng)態(tài)協(xié)同組合的兩層框架,組合過程完全由一些參與組合的服務(wù)Agent協(xié)同完成,保障服務(wù)組合的效率及準(zhǔn)確率。 (2)首先,區(qū)分基礎(chǔ)服務(wù)、結(jié)構(gòu)化服務(wù)和非結(jié)構(gòu)化服務(wù)。因?yàn)槔秒x線POMDP算法在全局范圍搜索組合策略,雖然搜索全面但搜索耗時(shí)。本文采用PBOL_POMDP解決基礎(chǔ)服務(wù)到結(jié)構(gòu)化服務(wù)的組合過程,引入偏好信息的離線POMDP算法,既處理了全局的最優(yōu)規(guī)劃問題,也增大了搜索的目的性,大大減小了搜索耗費(fèi)的時(shí)間,在長期偏好和中期偏好沒有變化之前,搜索結(jié)果作為重要的數(shù)據(jù)信息保存下來。 (3)在傳統(tǒng)在線POMDP算法之上,為了求解需求功能復(fù)雜的服務(wù)組合,將離線算法得到的解集合的上下界作為在線算法的啟發(fā)式信息,進(jìn)一步應(yīng)用到在線算法中。采用HO_POMDP服務(wù)組合算法解決基礎(chǔ)服務(wù)與前一步生成的所有服務(wù)之間的組合。 (4)搭建實(shí)驗(yàn)環(huán)境,將MADC_SCS框架和服務(wù)組合過程用JADE平臺(tái)模擬出來,將自己部署的簡單服務(wù)封裝實(shí)驗(yàn),將離線和在線POMDP算法擴(kuò)展后應(yīng)用到服務(wù)的動(dòng)態(tài)協(xié)同組合過程中,有效地改善了服務(wù)組合的成功率和時(shí)耗特性。
[Abstract]:In the era of rapid development of Internet technology and computer technology, Web services and their composition technologies have developed rapidly in both academia and industry. As an open mode of business delivery, Web services have been widely recognized as having excellent platform independence. In order to meet the needs of users, the methods of service recommendation and service composition appear rapidly. But in these processes, the services can only wait for the third party to call, and the number of services increases, making many services have not been invoked for a long time. The software Agent has the characteristics of initiative, autonomy and evolution, it can obtain the third party's requirement information, recommend itself to it, and make use of the cooperative technology to realize stronger function after making up the multi-Agent system. But it doesn't have the platform independence, interoperability, and flexibility that web services do. In order to make the service itself intelligent, the intelligent web service and the cooperative service composition algorithm are obtained, and the composition efficiency and the composition success rate of the service are verified. By combining Agent with web services, the service grabs the information of the users' needs in the network, and recommends itself to the service requesters in real time and actively. The Jade platform is used to deal with the dynamic composition. Agent encapsulated services using the preference based offline observable Markov decision Service composition algorithm (PBOLSP) and the heuristic online part. Observe that the Markov decision Service composition algorithm (HOSP) is efficient, Independently completes the process of service composition and returns the result to the service caller. The results are compared with the classical AEMS2 algorithm and the OPUMO1 algorithm in the citation, and the performance of the algorithm is verified. Based on this, this paper has done the following work: (1) introduce Agent to extend web services, extend the web service composition model in the Internet to multi-Agent dynamic cooperative service composition system, summarize the two-tier function structure of dynamic collaboration based on multi-Agent system. The framework of dynamic cooperative service composition system for performing service composition is presented. The traditional two-tier framework and the extended two-tier framework of multi-Agent dynamic cooperative composition are compared. The composition process is completed entirely by some participating service Agent cooperation. Ensure the efficiency and accuracy of service composition. (2) first, distinguish between basic services, structured services and unstructured services. Because offline POMDP algorithm is used to search the combination strategy in the global scope, the search is comprehensive but time-consuming. In this paper, PBOLPOMDP is used to solve the composition process from basic services to structured services, and an off-line POMDP algorithm with preference information is introduced. It not only deals with the global optimal programming problem, but also increases the purpose of search, and greatly reduces the time spent in searching. Search results are preserved as important data information before long-term and medium-term preferences have not changed. (3) in order to solve the complex service composition with complex requirements, search results are saved on the basis of traditional online POMDP algorithm. The upper and lower bound of the solution set obtained by the offline algorithm is regarded as the heuristic information of the online algorithm, which is further applied to the online algorithm. The HOP-POMDP service composition algorithm is used to solve the composition between the basic services and all the services generated in the previous step. (4) the experimental environment is built, and the MADCSCS framework and service composition process are simulated by Jade platform. The self-deployed simple service encapsulation experiment is applied to the dynamic cooperative composition process of services by extending the offline and online POMDP algorithms. The success rate and time-consuming characteristics of service composition are improved effectively.
【學(xué)位授予單位】：陜西師范大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TP393.09

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 章宗長;陳小平;;雜合啟發(fā)式在線POMDP規(guī)劃[J];軟件學(xué)報(bào);2013年07期

2 桂林;武小悅;;部分可觀測馬爾可夫決策過程算法綜述[J];系統(tǒng)工程與電子技術(shù);2008年06期

，

本文編號(hào)：2138000