本文選題：任務(wù)劃分 + 分布式負(fù)載平衡��； 參考：《電子科技大學(xué)》2014年碩士論文

【摘要】：隨著科技發(fā)展,相控陣?yán)走_(dá)負(fù)載任務(wù)越來越重,雷達(dá)系統(tǒng)需要處理的數(shù)據(jù)量急劇增加,計算難度增大,但系統(tǒng)對實時性的要求沒有降低。采用單個處理器處理雷達(dá)高負(fù)載任務(wù)已經(jīng)難以滿足相控陣?yán)走_(dá)任務(wù)實時性的需求。為了快速高效處理相控陣?yán)走_(dá)系統(tǒng)負(fù)載任務(wù),采用分布式系統(tǒng)并行處理思想,利用調(diào)度算法將雷達(dá)可并行任務(wù)指派到客戶機并行處理,降低任務(wù)完成時間,滿足雷達(dá)系統(tǒng)實時性要求;通過調(diào)度算法對任務(wù)合理調(diào)度,充分利用系統(tǒng)資源,提高分布式負(fù)載平衡率。據(jù)此,對相控陣?yán)走_(dá)任務(wù)、調(diào)度環(huán)境建立數(shù)學(xué)模型,優(yōu)化指派算法和粒子群算法,并將兩種算法分別應(yīng)用于任務(wù)調(diào)度中,研究任務(wù)的實時調(diào)度和負(fù)載分配技術(shù),具體研究工作如下:1.相控陣?yán)走_(dá)負(fù)載任務(wù)和分布式調(diào)度環(huán)境建模,確立兩者間映射關(guān)系。鑒于任務(wù)串行和并行處理差異,雷達(dá)任務(wù)劃分成通信開銷小、適于在分布式系統(tǒng)中調(diào)度的子任務(wù),并通過任務(wù)相關(guān)性體現(xiàn)子任務(wù)間的數(shù)據(jù)傳輸和通信。從任務(wù)相關(guān)性、復(fù)雜度和任務(wù)量三方面建立雷達(dá)負(fù)載任務(wù)數(shù)學(xué)模型。選擇“服務(wù)器-客戶機”主從式調(diào)度環(huán)境,分別對客戶機和通信網(wǎng)絡(luò)建立數(shù)學(xué)模型。確立相控陣?yán)走_(dá)負(fù)載任務(wù)在分布式異構(gòu)系統(tǒng)中調(diào)度的目標(biāo)函數(shù)和約束條件,確立任務(wù)與環(huán)境映射關(guān)系。2.提出一種基于指派問題和匈牙利算法的優(yōu)化指派算法。鑒于指派算法中指派策略導(dǎo)致客戶機空閑等待時間過長或通信消耗過大、匈牙利算法會陷入死循環(huán)問題,提出優(yōu)化指派算法:利用任務(wù)量作為指派標(biāo)準(zhǔn),平衡客戶機空閑時間和通信消耗之間的矛盾;以任務(wù)累積量與閾值之間大小比較作為是否指派任務(wù)到客戶機的標(biāo)準(zhǔn),調(diào)整系統(tǒng)負(fù)載平衡率;以任務(wù)完成時間和負(fù)載平衡率兩個指標(biāo)為目標(biāo),避免匈牙利算法陷入死循環(huán)。通過數(shù)值仿真和算法對比,驗證優(yōu)化算法避免了匈牙利算法陷入死循環(huán),縮短了任務(wù)完成時間,極大地改善了系統(tǒng)負(fù)載平衡。3.提出一種基于離散粒子群算法的優(yōu)化算法。針對粒子群算法收斂速度快但易陷入局部最優(yōu)解的缺陷,提出粒子群優(yōu)化算法:按效率矩陣轉(zhuǎn)化概率初始化粒子,增加粒子多樣性;通過迭代次數(shù)自適應(yīng)調(diào)整慣性系數(shù);將任務(wù)完成時間和負(fù)載平衡率的加權(quán)作為適應(yīng)度函數(shù),通過適應(yīng)度函數(shù)值負(fù)反饋給學(xué)習(xí)因子,雙重調(diào)整全局與局部搜索,減少算法陷入局部最優(yōu)解的風(fēng)險。數(shù)值仿真驗證了算法對收斂速度和避免算法陷入局部最優(yōu)解的有效性,減少了任務(wù)完成時間,提高了系統(tǒng)負(fù)載平衡率。4.系統(tǒng)實現(xiàn)。將優(yōu)化指派算法與粒子群優(yōu)化算法分別置于完整的雷達(dá)仿真系統(tǒng)中進(jìn)行仿真,驗證了算法在仿真系統(tǒng)中的優(yōu)化效果。
[Abstract]:With the development of science and technology, the task of phased array radar is becoming more and more heavy. The amount of data needed to be processed by radar system increases sharply, and the calculation difficulty increases, but the requirement of real-time performance is not reduced.It is difficult to meet the real-time requirement of phased array radar task by using a single processor to deal with high-load task.In order to deal with the phased array radar system load task quickly and efficiently, the distributed system parallel processing idea is adopted, and the radar parallel task is assigned to the client parallel processing by using the scheduling algorithm, which reduces the task completion time.It can meet the real-time requirements of radar system and make full use of system resources to improve the distributed load balancing rate.Based on this, the mathematical model of phased array radar task scheduling environment is established, the assignment algorithm and particle swarm optimization algorithm are optimized, and the two algorithms are applied to task scheduling, and the real-time task scheduling and load allocation techniques are studied.The specific research work is as follows: 1.Based on the modeling of phased array radar load task and distributed scheduling environment, the mapping relationship between them is established.Due to the difference between serial and parallel processing, radar tasks are divided into sub-tasks with low communication overhead, which are suitable for scheduling in distributed systems, and the data transmission and communication among sub-tasks are reflected by task correlation.The mathematical model of radar load task is established from three aspects: task correlation, complexity and task quantity.Selecting the "server-client" master and slave scheduling environment, the mathematical models of client and communication network are established.The objective function and constraint condition of phased array radar load task scheduling in distributed heterogeneous system are established, and the mapping relation between task and environment is established.An optimal assignment algorithm based on assignment problem and Hungarian algorithm is proposed.In view of the fact that the assignment strategy in the assignment algorithm results in too long idle waiting time or too much communication consumption, the Hungarian algorithm will fall into a dead-loop problem. This paper proposes an optimized assignment algorithm, which uses the amount of task as the assignment criterion.Balancing the contradiction between client idle time and communication consumption, adjusting system load balancing rate by comparing the size of task accumulation and threshold as the criterion of whether to assign task to client;Aiming at task completion time and load balancing rate, the Hungarian algorithm is avoided from falling into a dead cycle.Through numerical simulation and algorithm comparison, it is verified that the optimization algorithm avoids Hungary algorithm from falling into dead-cycle, shortens the time of task completion and greatly improves the system load balance. 3.An optimization algorithm based on discrete particle swarm optimization (DPSO) is proposed.Aiming at the defect of particle swarm optimization (PSO), which converges fast but is easy to fall into local optimal solution, particle swarm optimization (PSO) algorithm is proposed: initializing particles according to efficiency matrix transformation probability, increasing particle diversity, adjusting inertia coefficient adaptively by iterating times;The weighting of task completion time and load balancing rate is taken as fitness function. By negative feedback of fitness function value to learning factor, global and local search are adjusted to reduce the risk of the algorithm falling into local optimal solution.Numerical simulation shows that the algorithm is effective to convergence speed and avoid the algorithm falling into local optimal solution, reduces the task completion time and improves the load balancing rate of the system.System realization.The optimization assignment algorithm and particle swarm optimization algorithm are simulated in the complete radar simulation system respectively, and the optimization effect of the algorithm in the simulation system is verified.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TN958.92

【參考文獻(xiàn)】

相關(guān)期刊論文 前4條

1 蘭舟;孫世新;;基于動態(tài)關(guān)鍵任務(wù)的多處理器任務(wù)分配算法[J];計算機學(xué)報;2007年03期

2 尚亞靈;袁道華;丁瑩;曾祥洪;;消息傳遞機制的比較及其最新發(fā)展[J];計算機工程與設(shè)計;2007年10期

3 彭文敏;胡書;張莉;譚周田;;基于粒子群算法的無人機航跡規(guī)劃問題[J];中國高新技術(shù)企業(yè);2010年04期

4 馮冬青;楊書顯;;氧樂果合成過程的PSO-回歸BP網(wǎng)絡(luò)建模方法[J];鄭州大學(xué)學(xué)報(理學(xué)版);2011年03期

相關(guān)碩士學(xué)位論文 前2條

1 王延玲;基于MPICH的并行計算系統(tǒng)負(fù)載平衡技術(shù)的研究[D];曲阜師范大學(xué);2009年

2 朱麗;基于不確定時間下的MPBP生產(chǎn)調(diào)度算法的研究[D];江蘇科技大學(xué);2012年

，

本文編號：1749187