【作者】 靳肖闪；

【导师】 景宁；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2009， 博士

【摘要】 卫星对地成像是获取地球表面信息的一种重要手段。当前社会经济和军事等领域的对地成像应用体现出成像区域分布广泛、观测频繁、成像方式多样、成像时效性强等特点,带来对地成像需求的快速增长。随着以多种类型成像卫星、地面接收站和中继卫星系统等组成的对地成像系统的发展,作为对地成像系统控制核心的控制中心需要从全局角度分配成像卫星和接收资源来进行星地资源的综合调度,以最大程度满足对地成像需求。然而,目前卫星对地成像调度研究大多集中在成像阶段的调度或接收阶段的调度,缺乏将成像过程和接收过程进行整体优化考虑的方法和技术。论文提出了星地资源综合调度的思想来解决卫星成像过程和数据接收过程的整体优化问题,以星地综合调度模型为基础,重点研究了基于置换表示的调度优化技术,并通过松弛优化方法验证了所提方法的有效性。主要工作和创新点包括:1、建立了包含多类型星地资源的星地综合调度数学模型。星地综合调度模型是星地综合调度研究的基础。论文分析了成像卫星和接收资源的工作特性,研究了星地综合调度问题的组成要素及其相互关系,提出了将成像活动和接收活动进行组合的星地综合调度数学模型,并可适应调度周期的变化。2、提出了基于置换表示的星地综合调度问题可行解表示方法。星地综合调度问题是非线性复杂约束优化问题,直接在问题空间上进行搜索时可行调度的构造和优化难以实现。为此,论文研究了基于置换表示的可行解表示方法,通过为置换序列分配星地资源,将可行调度优化转化为无约束空间上的搜索问题。3、研究了两类可行调度优化算法。最优调度的搜索由在置换空间上搜索具有最大评价值置换序列的过程实现。论文基于邻域图和对称群理论,分析了置换空间上交换邻域和插入邻域的性质,提出了具有简单结构的有记忆随机邻域搜索算法,可适应对调度优化时间的需要。为了获得更优化的调度,论文提出了一种混合遗传算法,在保留全局收敛性的同时,以邻域搜索算法增强了遗传算法的局部寻优能力。4、研究了基于拉格朗日松弛的星地综合调度可行解优化评价方法。星地综合调度是一类NP难解的组合最优化问题,优化算法只能在可接受花费下给出该问题的近似最优解,但所得可行解的优化程度需要其它优化方法来评价。论文基于具有多项式复杂度的最长路径算法和次梯度优化算法求解星地综合调度松弛问题,以获得原问题的紧致上界。实验结果显示,基于置换表示方式的可行解优化算法,所得可行解非常接近松弛方法所得上界,证明了论文提出的星地综合调度优化方法具有较高的优化能力,可以满足星地综合调度问题的优化需求。基于以上研究成果,论文最后设计并实现了星地综合调度实验系统,并应用于一类包含多颗成像卫星和多个地面站的全球目标观测应用,验证了所提出技术和方法的有效性。更多还原

【Abstract】 Imaging from space is an important way to obtain the information about the earth’s surface. Nowadays, the imaging applications in the field of social economy and military operation are featured by: broadly distributed imaging areas, frequent observing, various kinds of imaging satellites, instant obtaining of the imaging data, all of which lead to the rapid proliferation of earth observing demands. With the development of earth observing system composed of various imaging satellites, ground stations and data relay satellites, the control center as the core of the earth observing system needs to globally assign tasks to different imaging satellites and the corresponding receiving networks in order to image and receive image data, so that the earth observing needs can be maximally met. However, most of the current researches on earth observing scheduling are classified into two independent categories, i.e., satellites imaging scheduling and imaging data receiving scheduling. Thus there lacks a holistic method to integrate the two stages.This dissertation proposes the integrated scheduling techniques for imaging satellites to holistically optimize both of the imaging process and data receiving process. In the dissertation, the integrated scheduling model is taken as the research foundation; scheduling optimization technologies through solution space transferring are taken as the research focus; moreover, relaxation methods are studied to validate the effectiveness of the proposed methods. The main work and contributions of this dissertation can be concluded as the following four parts:1. Building the integrated scheduling model that comprises various kinds of resources.The integrated scheduling model is the foundation of the whole research. Upon the thorough analysis on imaging satellites and the receiving network, the factors and their relations that should be considered in modeling the problem are summarized. An integrated scheduling model that comprises both the imaging process and receiving process is presented. Meanwhile, the proposed model is adaptive to changes of scheduling period.2. Putting forward solution transfer method by permutation based representation for the integrated scheduling.The integrated scheduling problem, which is non-linear and complicated, belongs to constraints optimization problems. Therefore, it is difficult to construct and optimize the feasible schedules through directly searching in the space of schedules. Thus, a solution space transfer method by permutation based representations is proposed. Through assigning resources to the permutations, the method transfer the searching of feasible scheduling from the space of schedules to the non-constraint space of permutations.3. Presenting two kinds of optimization algorithms in the space of permutations to obtain optimal schedules.Optimal schedules can be obtained by searching in the space of permutations for individuals with maximum fitness value. Characteristics of the swap neighborhood and insertion neighborhood in the space of permutations are analyzed based on neighborhood graph and symmetric group theories. Then the dissertation puts forward a stochastic neighborhood search algorithm with memory, which has simple structure and converges very fast. To obtain more optimized schedules, a hybrid genetic search algorithm is proposed. The algorithm can keep its global astringency; what’s more, it strengthens its local optimizing ability through neighborhood searching.4. Presenting Lagrange relaxation method to estimate the optimized integrated scheduling.The Integrated scheduling problem belongs to NP-Hard combinatorial optimization problems. The search algorithms can only approach the optimize schedules at limited computing costs. Thus, methods are needed to evaluate the optimal schedules. The dissertation proposes Lagrange relaxation method to solve this problem. A longest path search algorithm with polynomial complexity and a subgradient optimization algorithm are presented to solve the relaxation problem; therefore, the tight upper bound for the original problem can be gained. The experimental results show that schedule from the search algorithms in the space of permutations is very close to the upper bound, which proves that our proposed optimization algorithms are effective and can meet the needs of optimizing the integrated scheduling.Based on the above achievements, the thesis designs and implements an experimental system, which has been used to solve a worldwide earth observation proble through multiple imaging satellites and receiving groundstations. This system validates the efficiency and practicability of the presented techiques.更多还原