求解多星任务规划问题的演化学习型蚁群算法  

A learnable ant colony optimization to the mission planning of multiple satellites

作　　者：陈英武[1] 姚锋[1] 李菊芳[1] 贺仁杰[1] 邢立宁[1] 

机构地区：[1]国防科技大学信息系统与管理学院管理系,长沙410073

出　　处：《系统工程理论与实践》2013年第3期791-801,共11页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(71031007;71101150;71071156;61203180;71101013)

摘　　要：任务规划在成像卫星指挥控制过程中起着非常关键的作用,在成像卫星应用系统中处于神经中枢的地位.提出了一种求解多星任务规划问题的演化学习型蚁群算法:在参数绩效知识的指导下,采用动态参数模型为下次迭代随机选择较为合理的参数组合;从优化过程中不断地抽取构件知识,采用构件知识指导人工蚂蚁在后续优化过程中构建可行方案.在蚁群算法、动态参数决策模型和构件知识的共同作用下,演化学习型蚁群算法的优化绩效得到了极大提高.采用多星任务规划问题的21个测试实例进行实验,结果表明演化学习型蚁群算法在优化性能方面优于其他两种方法.Mission planning plays a very important role in the management process of imaging satellites, and it is the hardcore of imaging satellites application systems. A learnable ant colony optimization （LACO） is proposed to the mission planning of multiple satellites. Before each iterative loop, the LACO randomly selects an appropriate parameter combination via dynamic parameter decision model according to the performance knowledge of parameters. Different than standard ant colony optimization, the LACO extracts some available component knowledge, and applies the obtained component knowledge to guide artificial ants to construct feasible solutions in the subsequent optimization process. Under the effective cooperation of ant colony optimization, dynamic parameter decision model and component knowledge, the performance of LACO was largely improved. Twenty-one testing instances were applied to compare the performance of different approaches. Experimental results suggest that the LACO outperforms other two approaches.

关 键 词：任务规划 参数绩效知识 参数动态调整 构件知识 蚁群算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]