对地观测卫星系统部署-调度一体化问题研究

Earth observation satellite system (EOSS) works as the main space platform acquiring earth images. Its observation performance is determined by two coupled problems of system configuration and scheduling. Problem of system configuration is solved to improve EOSS performance by optimizing orbit and sensor parameters of satellites. Problem of scheduling is to maximize the number of satisfied user requests by reasonably allocate satellites. The two problems are interactively coupled by common variables influencing the observation efficiency of EOSS. EOSS of our country is of low efficiency due to the separation of two problems. Therefore, problem of integrated configuration and scheduling of EOSS is investigated. For regular situation it is taken as a high-dimensional black-box optimization problem in continuous space which is solved by a surrogate model based approximation method. For emergency situation, the problem becomes a bi-level optimization problem. A combined framework of genetic algorithm (GA) and adaptive large neighborhood search (ALNS) algorithm is proposed to solve it. GA is used to generate different system configurations which will work as inputs of EOSS scheduling problem. ALNS will create a schedule so that different satellites can work in a collaborated way. Then the created schedule is fed back to GA as its fitness merit to judge whether the configuration is optimized or not. The acquisitions of this project raise a new method to solve the bi-level optimization problem. At the same time they provide important theoretical supports for the development of future EOSS in our country.

对地观测卫星系统(EOSS)是当前最主要的太空信息获取平台，其观测效能取决于系统的部署和调度两个方面。部署子问题通过优化卫星的轨道和载荷配置参数实现对系统性能的改进，调度子问题在既定部署方案下通过合理分配和安排卫星资源实现对用户需求的最大化满足，两者互相耦合、互相影响。当前，我国EOSS的部署与调度相互分离，致使系统效能不高。针对此，本项目研究EOSS部署-调度的一体化问题，常规模式下将问题建模为连续空间的高维黑箱优化问题，提出基于代理模型的近似优化方法进行求解；应急模式下将问题建模为双层优化问题，构建了基于遗传算法（GA）和自适应大邻域搜索（ALNS）算法的求解框架，GA用以生成系统的部署方案并将其作为调度问题的输入，ALNS用以实现不同卫星的协同调度，调度结果作为GA适应度函数对部署方案进行评估。本项目提供了一种新的双层优化问题的求解框架，对我国未来对地观测卫星系统的发展具有重要的理论支持。

我国对地观测卫星系统的研制建设与管控分属不同部门，这种体制上的约束导致了EOSS的部署优化与规划调度相分离：在系统建设过程中仅考虑系统性能指标的优化，没有考虑实际应用情况；在卫星日常管控中，由于系统部署的限制，使得系统对很多任务的完成效果不佳，因而使得整个系统的效能发挥受到限制。针对此本项目提出了对地观测卫星系统部署-调度一体化问题，其中部署子问题通过优化卫星的轨道和载荷配置参数实现对系统性能的改进，调度子问题在既定部署方案下通过合理分配和安排卫星资源实现对用户需求的最大化满足，两者互相耦合、互相影响。. 为了确保求解框架的适用性，本项目针对卫星应用的不同场景，将其分为常规模式和应急模式。常规模式下将问题建模为连续空间的高维黑箱优化问题，并提出基于代理模型的近似优化方法进行求解，为了确保模型的精度，本课题采用了基于多点更新的代理模型构建机制，并采用基于模式搜索的方法，快速搜索代理模型的更新点；应急模式下将问题建模为双层优化问题，上层解决卫星部署问题，下层解决卫星调度问题，构建了基于遗传算法和自适应大邻域搜索算法相结合的求解框架，遗传算法采用变长编码方式，用以生成系统的部署方案并将其作为调度问题的输入，自适应大邻域搜索算法用以实现不同卫星的协同调度，尤其是针对敏捷卫星，设计了时间松弛机制以缓解观测开始时间的时间依赖特性所带来的求解难度，调度结果作为GA适应度函数对部署方案进行评估。最后我们通过多个应用实例证明了项目所提算法框架的有效性。. 本项目所提出的敏捷卫星任务规划算法已经成功应用在在轨卫星的任务规划中，求解效果得到用户的认可，本项目提出的双层优化框架对我国未来对地观测卫星系统的发展论证和建设规划具有重要的理论支持。

内容获取失败，请点击重试