【摘要】：從以話音業(yè)務(wù)為代表的第一代模擬通信系統(tǒng)到如今大規(guī)模商用的第四代(4G)移動通信系統(tǒng),移動通信網(wǎng)絡(luò)走過了漫長歷程。在移動通信網(wǎng)絡(luò)中實現(xiàn)移動云計算、移動多媒體等新興業(yè)務(wù)已經(jīng)成為移動通信的下一步演進目標。當下,隨著智能終端的迅速普及和網(wǎng)絡(luò)通信量的爆炸性增長,面向第五代(5G)移動通信的無線通信技術(shù)已經(jīng)吸引了業(yè)內(nèi)人士的極大關(guān)注。D2D(Device-to-Device:D2D)通信作為5G的關(guān)鍵候選技術(shù),已經(jīng)成為一個研究熱點。因為蜂窩網(wǎng)絡(luò)引入D2D通信技術(shù)后,網(wǎng)絡(luò)和用戶的資源分配會發(fā)生巨大的變化,所以資源控制的優(yōu)化設(shè)計就成為一個研究重點。資源控制包含模式選擇、資源分配和功率控制,為了使網(wǎng)絡(luò)達到最佳性能,越來越多的研究工作把這三個機制放在一起研究。目前,多數(shù)針對D2D通信的研究都是基于無限積壓業(yè)務(wù)模型和分組級業(yè)務(wù)模型,而本論文基于更適合于新一代無線通信的流級業(yè)務(wù)模型,主要對資源控制當中的模式選擇和資源分配進行優(yōu)化設(shè)計。在引入D2D通信的蜂窩網(wǎng)絡(luò)中應(yīng)用正交頻分多址技術(shù)(Orthogonal Frequency Division Multiple Access:OFDMA),研究該蜂窩網(wǎng)絡(luò)的資源控制優(yōu)化問題,從而使網(wǎng)絡(luò)中流傳輸?shù)钠骄芰肯淖钚�。本論文以排隊�?QueuingTheory)為基礎(chǔ),將上述最優(yōu)化問題建立成無限范圍平均回報的馬爾可夫決策過程(Infinite Horizon Average Reward Markov Decision Process)模型。經(jīng)典的馬爾可夫決策過程模型求解方法為貝爾曼方程(Bellman's Equation),即傳統(tǒng)集中式離線的值迭代算法。為了解決求解馬爾可夫決策過程模型所面臨的"維數(shù)災(zāi)難"問題,本論文把貝爾曼方程簡化成等效貝爾曼方程(Equivalent Bellman's Equation)。建立Q值函數(shù)與等效貝爾曼方程中值函數(shù)的聯(lián)系,同時利用線性近似(Linear Approximation)方法對全局Q值函數(shù)做進一步化簡,并應(yīng)用在線隨機學(xué)習(xí)(Online Stochastic Learning)算法更新迭代Q值函數(shù)。以排隊論、馬爾可夫決策過程為理論基礎(chǔ),本論文提出將等效貝爾曼方程、線性近似方法和隨機在線學(xué)習(xí)算法三者結(jié)合的分布式資源控制算法,用以優(yōu)化蜂窩網(wǎng)絡(luò)中D2D通信的模式選擇和資源分配方式,從而達到網(wǎng)絡(luò)中流傳輸?shù)钠骄芰肯淖钚〉膬?yōu)化目標。搭建仿真平臺將本論文提出的算法與其他四種模式選擇-資源分配算法相對比,仿真結(jié)果顯示采用提出算法的網(wǎng)絡(luò),其流傳輸?shù)钠骄芰肯淖钚　?br/>[Abstract]:The mobile communication network has come a long way from the first generation analog communication system represented by voice service to the fourth generation (4G) mobile communication system which is now large-scale commercial. Mobile cloud computing, mobile multimedia and other emerging services in mobile communication networks have become the next evolution goal of mobile communication. At present, with the rapid popularization of intelligent terminals and the explosive growth of network traffic, The wireless communication technology for the fifth generation (5G) mobile communication has attracted great attention of the industry. D2D (Device-to-Device:D2D) communication as a key candidate technology of 5G has become a research hotspot. With the introduction of D2D communication technology in cellular networks, the resource allocation of network and users will change greatly, so the optimization design of resource control becomes a research focus. Resource control includes mode selection, resource allocation and power control. In order to achieve optimal network performance, more and more research work put these three mechanisms together. At present, most of the research on D2D communication is based on infinite backlog service model and packet-level traffic model, but this paper is based on the flow level service model which is more suitable for the new generation wireless communication. The mode selection and resource allocation in resource control are optimized. The orthogonal frequency division multiple access (Orthogonal Frequency Division Multiple Access:OFDMA) technique is applied in the cellular network with D2D communication. The resource control optimization problem of the cellular network is studied, so that the average energy consumption of the network is minimized. On the basis of queue theory (QueuingTheory), the above optimization problem is established as a Markov decision process (Infinite Horizon Average Reward Markov Decision Process) model with infinite range average return. The classical Markov decision process model is solved by the Belman equation (Bellman's Equation), which is the traditional centralized off-line value iterative algorithm. In order to solve the problem of "dimensionality disaster" in solving the Markov decision process model, this paper simplifies the Berman equation into the equivalent Belman equation (Equivalent Bellman's Equation). The relationship between Q value function and the median function of equivalent Berman equation is established. The global Q value function is further simplified by linear approximate (Linear Approximation) method, and the iterative Q value function is updated by online random learning (Online Stochastic Learning) algorithm. Based on queuing theory and Markov decision process, this paper proposes a distributed resource control algorithm which combines equivalent Berman equation, linear approximation method and stochastic online learning algorithm. It is used to optimize the mode selection and resource allocation of D2D communication in cellular networks, so as to achieve the goal of minimum average energy consumption in the network. The simulation platform is built to compare the proposed algorithm with the other four mode selection-resource allocation algorithms. The simulation results show that the proposed algorithm has the lowest average energy consumption.
【學(xué)位授予單位】：北京交通大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TN929.5

【參考文獻】

相關(guān)期刊論文 前2條

1 焦巖;高月紅;楊鴻文;楊大成;;D2D技術(shù)研究現(xiàn)狀及發(fā)展前景[J];電信工程技術(shù)與標準化;2014年06期

2 王俊義;鞏志帥;符杰林;陳小徽;林基明;;D2D通信技術(shù)綜述[J];桂林電子科技大學(xué)學(xué)報;2014年02期

相關(guān)碩士學(xué)位論文 前2條

1 鄺奕如;蜂窩網(wǎng)絡(luò)D2D技術(shù)中模式選擇機制的優(yōu)化設(shè)計[D];北京交通大學(xué);2016年

2 趙靜;LTE-A蜂窩網(wǎng)絡(luò)中D2D通信資源分配機制的流級建模與性能分析[D];北京交通大學(xué);2013年

，

本文編號：2342904