摘要
以目标跟踪为背景,研究了单平台上主被动传感器的长期调度问题。通过合理、实时地切换主被动传感器,使得有限时域内的跟踪精度和辐射风险达到合理的平衡。将该调度问题构建成部分可观马氏决策过程(partially observable Markov decision process,POMDP)以同步实现目标跟踪和辐射控制。提出以容积采样法估算长期精度收益,以隐马氏模型滤波器推导长期辐射代价。最终将原问题转化成决策树并利用分枝定界法进行求解。仿真结果证明了本方法的有效性。
A non-ruyopic scheduling problem of how to intelligently and dynamically select active/passive sensors in a single platform for target tracking is studied. The goal is to develop a non-myopic scheme to make an optimal trade-off between the tracking accuracy and the radiation risk in a period of time. The problem is for mulated as a partially observable Markov decision process (POMDP) to solve target tracking and emission con- trol together. The cubature sampling is proposed to evaluate the long-term accuracy reward, and the long-term radiation cost is derived from the hidden Markov model filter. The original problem is transformed into a deci sion tree and solved using the branch and bound method. Simulation results demonstrate the effectiveness of the proposed scheduling anoroach.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2014年第3期458-463,共6页
Systems Engineering and Electronics
基金
军内科研重点项目资助课题
关键词
长期调度
部分可观马氏决策过程
决策树
分枝定界
non-myopic scheduling
partially observable Markov decision process (POMDP)
decisiontree
branch and bound