复杂战场环境下的几个目标跟踪新问题研究

Modern battlefield environment has changed significantly with the development of a variety of advanced military sensors. As a result, many military systems have specified more complicated requirements for target tracking. Because of the emergence of advanced sensors, some new challenging tracking problems, which are significantly different from traditional tracking problems, have appeared. Therefore, existing methods are not suitable for current tracking tasks. In order to solve the new and challenging target tracking problems in complicated battlefield environment, we propose the recently developed random finite set (RFS) theory and finite set statistics (FISST) approach as the main research tool, which will be complemented with the Dempster-Shafer (D-S) evidence theory, feedback fusion, intelligent optimization, learning algorithms, and compressed sensing theory, et al. This project will focus on: (1) new approaches for group target and extended target tracking using single or multiple sensors, (2) joint detection, recognition and tracking for stealth and high maneuvering targets, (3) targets tracking in unknown clutter, (4) integrated methods for "non-cooperative" targets tracking and moving platform multisensor spatial registration, and (5) performance evaluation methods and verification of system design. Each of the above five key scientific problems has specific background and application. However, the research methods of these problems are closely interrelated and mutually supportive. Research in this project has not only important academic and real-world problem-solving merits, but also has practical significance for improving the defense and attack abilities of our military system.

随着各种先进军用传感器的发展，现代战场环境发生了显著变化，各种应用系统对目标跟踪提出了日益复杂的要求，导致一些与传统跟踪问题相区别的新型跟踪问题出现，而现有的方法已不能胜任当前的跟踪任务。 本项目以复杂战场环境下的目标跟踪新问题为研究对象，以随机有限集和有限集合统计方法为主要研究工具，结合D-S证据理论、反馈融合方法、智能优化和学习算法以及压缩感知理论等，重点研究：（1）基于单或多传感器的群目标和扩展目标跟踪新方法；（2）隐身目标和高机动目标联合检测、识别和跟踪；（3）未知杂波环境下的目标跟踪；（4）"非合作"目标跟踪和动平台多传感器空间配准一体化方法；（5）性能评估方法和验证系统设计。这五个关键科学问题在研究内容上分别有各自明确的背景和应用，但在研究方法上紧密联系、相互支持。 本项目的研究不仅具有重要的学术价值和应用前景，而且对于提高我国国防系统的防御和攻击能力具有重要的现实意义。

本项目针对负责战场环境下的多目标跟踪新问题展开研究，具体研究内容包括：1）群目标和扩展目标跟踪算法研究；2）多机动目标跟踪算法研究；3）多目标联合检测与估计的误差界研究；4）多目标滤波器的收敛性分析研究；5）基于多目标均方误差界的多传感器控制算法研究。取得的重要结果包括：首先定义了标签分配(LA)距离，并证明了当阶数大于等于1时它为一个测度；其次得到了基于标签分配测度的多目标估计误差下界；并通过高斯混合给出了该下界的解析形式。最终，将上述理论应用于如下三个具体的传感器管理场景：A、多平台多传感器位置控制；B、分散式大规模网络的传感器选择；C、障碍环境下无线可移动传感器网络的路径规划。..本研究成果的科学意义为：在现代战场环境下，随着目标机动性、隐身性和远距巡航能力的提高；目标编队或协同作战使得目标密集程度增加；电子对抗等干扰技术的进步；复杂的地面或海面环境等各种因素的共同影响，使得单个传感器已无法满足跟踪要求。因此，迫切需要利用本项目所建议的多目标跟踪和传感器管理方法使多个传感器平台协同配合，共同完成跟踪任务。综上，本项目的研究在军用和民用领域都有着巨大的应用潜力。