基于子模优化的远程预警传感器管理研究

项目名称： 基于子模优化的远程预警传感器管理研究

项目编号： No.61503305

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 其他

项目作者： 王增福

作者单位： 西北工业大学

项目金额： 18万元

中文摘要： 远程预警等一类大规模感知系统的传感器资源管理与调度面临着NP-Hard、决策空间与状态空间的高维数等问题。传统最优算法难以在系统要求时间内给出最优解，传统近似方法无法从理论上获得算法性能下界及运行时间的定量描述与预估，难以满足系统大规模、高可靠、强实时等要求。近年来取得重要突破的子模优化为解决该类问题提供了新途径，能够获得多项式时间计算复杂度的最优解或具有性能下界的近似解。本项目拟围绕子模优化理论，针对远程预警多传感器动态配置部署、传感器-目标分配、资源自动调度等问题，构造具有子模性质的优化目标函数，研究可行解空间的组合结构；在此基础上，采用归约法归证待优化问题的计算复杂性，建立具有性能保证的多项式时间计算复杂度远程预警传感器管理快速方法；并在天波超视距雷达网典型数据下进行仿真验证。本项目拟为远程预警资源管理与调度提供新原理和新方法的同时，推动子模优化在快速算法和近似保证等方面的发展。

中文关键词： 多传感器组网；多平台协同探测；闭环系统融合与控制

英文摘要： National early-warning system is a large scale sensing system. The sensor resources management and scheduling of such kind of systems are faced with NP-Hardness and high dimensional action space and state space. Traditional optimal methods are hard to obtain solutions in a reasonable time. Approximate algorithms in literature are unable to provide the theoretical lower bound of their performance and the evaluation and prediction of their running times. As a result, they are hard to satisfied system’s demands of large scale, high reliability, and real time. Submodular function optimization makes breakthrough recently, and provides a new way, since algorithms developed based on it can obtain an optimal solution or approximate solution with performance guarantee in polynomial time. To the problems of multi-sensor dynamic deployment and placement, sensor-target assignment, and resources scheduling for early warning, based on submodular optimization, we will firstly develop submodular objective functions and explore the combinatorial structure of the space of feasible solutions. Then the computational complexity of the proposed problem will be proved by reduction. Finally we will develop fast and polynomial time algorithms with performance guarantee for sensor management in early-warning system, and then verify them in simulated over-the-horizon radar network with typical data. In this project, we will develop novel high efficient algorithms for large scale and complex optimization problems, and contribute to theory and methods for early warning systems. Meanwhile, this project will promote the development of submodular function optimization in terms of fast algorithms and performance guarantee.

英文关键词： Sensor Network;Multiplatform Cooperative Sensing;Closed Loop system fusion and control