性能驱动的无线传感网目标跟踪动态传感器选择策略

项目名称： 性能驱动的无线传感网目标跟踪动态传感器选择策略

项目编号： No.61473047

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 其他

项目作者： 杨小军

作者单位： 长安大学

项目金额： 81万元

中文摘要： 对能量和带宽等资源受限的无线传感器网络，有效的传感器选择策略是实现分布式目标跟踪的关键技术。本项目拟从传感器选择准则入手，研究多步、多目标跟踪的条件后验克拉美-罗下界（PCRLB）及其分布式实现，建立一种计算有效的传感器最优性能的量化测度。以条件PCRLB作为回报函数，研究非近视（Nonmyopic）传感器选择策略，基于Rollout策略迭代算法，寻找计算有效的Nonmyopic次优策略。对检测不确定、关联不确定、参数不确定和个数不确定的多目标跟踪问题，基于随机有限集理论建立传感器管理模型，研究性能驱动的多目标跟踪传感器选择策略。对存在拓扑变动和链路故障的分散式结构的传感器网络，基于Consensus方法和Gossip方法，研究分布式传感器节点选择策略及算法。开展特色鲜明的能量、带宽和计算有效的多步多目标分布式多传感器动态选择策略的理论研究与算法实现，并完成相应的仿真分析和系统集成。

中文关键词： 目标跟踪；传感器管理；无线传感器网络；分布式估计；决策与优化

英文摘要： The efficient sensor selection scheme is the key technology to realize distributed target tracking for resource, such as energy and bandwidth and so on, constrained wireless sensor networks. Taking the criterion of sensor selection as the point of departure, we will investigate the multi-step-ahead conditional posterior Cramér-Rao lower bound (PCRLB), its extension to multi-target tracking and its distributed implementation. The project will develop the computation efficient quantized measure for the sensor achievable optimal performance. Further, by taking the conditional PCRLB as the reward function, the nonmyopic sensor selection approach will be explored. We will pursue the computation efficient nonmyopic sub-optimal scheme based on rollout policy iteration algorithm. In addition, for multi-target tracking in the presence of detection uncertainty, association uncertainty, parameter uncertainty and number uncertaity, we will establish the model of sensor management based on the random finite set theory. The project will seek the performance-driven sensor selection strategy for multi-target tracking. Also, for frequent changes of the network topology and link failure that exist in the decentralized senor networks, the distributed sensor selection scheme and algorithm will be investigated based on consensus and gossip algorithm. This project aims to carry out the distinct research on theory and algorithms for dynamic sensor selection strategies. The developed strategies are energy, bandwidth and computation efficient with the characteristics of multi-step-ahead, multi-target and distributed multi-sensor selection. Finally, we will complete the corresponding simulations and system integration.

英文关键词： Target Tracking;Sensor Management;Wireless Sensor Networks;Distributed Estimation;Decision and Optimization