摘要
针对高杂波、电子干扰环境,在量测驱动的多目标滤波框架下提出了一种基于决策不确定性的传感器管理方法。首先,根据部分可观测马尔科夫决策过程的理论,给出了基于Rényi信息增量的传感器管理一般方法。其次,综合考虑决策过程的信息完整性、信息质量、信息的内涵等因素,在量测驱动的自适应滤波框架下,基于目标运动态势评估多目标决策不确定性水平,并选取最大决策不确定性目标。最后,以最大决策不确定性目标的信息增量最大化为准则进行传感器分配方案的求解。仿真实验表明所提方法能够有效抑制电子干扰、杂波对多目标跟踪及传感器分配的影响,与基于威胁的传感器管理方法相比,所提方法的平均最优子模式分配(OSPA)距离及平均计算时长均显著降低,且在高杂波、电子干扰情形下具有较高的可靠性。
For electronic countermeasures and dense clutter environments,a sensor management algorithm based on decision uncertainty using the measurement-driven multi-target filter is proposed.First,according to the theory of partially observable Markov decision process,ageneral sensor management approach based on Rényi divergence is presented.Meanwhile,taking into account the information integrity,information quality and information connotation in the decision-making process,we evaluate the multi-target decision uncertainty level based on the target motion situation in the measurement-driven adaptive filtering framework,subsequently selecting the maximum decision uncertainty target.Finally,the sensor allocation scheme is solved with the maximum information gain of the maximum decision uncertainty target as the criterion.The simulation results show that the proposed algorithm can effectively suppress the influence of electronic countermeasures and dense clutter on multi-target tracking and sensor management.Compared with the threat-based sensor management algorithm,the average Optimal Sub-Pattern Assignment(OSPA)distance and the average calculation time are significantly reduced.In cases of dense clutter and electronic countermeasures,the proposed algorithm has high reliability.
作者
田晨
裴扬
侯鹏
赵倩
TIAN Chen;PEI Yang;HOU Peng;ZHAO Qian(School of Aeronautics,Northwestern Polytechnical University,Xi'an 710072,China;Science and Technology on Electro-Optic Control Laboratory,Luoyang 471023,China)
出处
《航空学报》
EI
CAS
CSCD
北大核心
2020年第10期262-275,共14页
Acta Aeronautica et Astronautica Sinica
基金
航空科学基金(20185153032)。
关键词
传感器管理
多目标跟踪
战术重要性标绘
量测驱动
部分可观马尔科夫决策过程
sensor management
multi-target tracking
tactical significance map
measurement-driven
partially observable Markov decision process