【作者】 李莹莹；

【导师】 刘云辉；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2008， 博士

【摘要】 本文研究基于移动无线传感器网络的目标跟踪问题,即利用传感器节点的感知、通信、计算和运动特性,对被探测区域内的未知运动目标进行连续跟踪。移动无线传感器网络是由大量可移动的传感器节点所组成的无线网络。它最显著的特点是节点资源有限,且抗毁性能很差。因此节能和分布式是必须考虑的因素。课题目的是提出一种分布式自主目标跟踪方案,在各种复杂环境下(如噪声、障碍物、扩散目标、多目标等),对目标保证高跟踪精度的同时,尽量最小化节点运动能耗,并且优化网络配置,包括始终保持网络连通、避免节点间以及节点与障碍物间发生碰撞等。本文分三个部分解决以上问题。第一部分研究在理想状态下节能跟踪所能达到的最优效果。首先证明节能跟踪问题的NP完全特性,然后提出一种类似贪心算法的局部中心式引导跟随算法,并且证明该算法的节点运动能耗在数值上逼近最小能耗,将其作为节能评价的标准。第二部分研究更具实际意义的分布式节能跟踪算法。首先建立分布式系统模型,规划节点间的信息交换机制,使得每时刻仅有少数节点处于活跃状态,自主运动来跟踪目标,其它节点均保持空闲。然后给出活跃节点的运动策略,包括三个模块:ⅰ)跟踪质量函数反映网络的跟踪性能,ⅱ)连接势函数反映网络的连通状态,ⅲ)分布式节能运动策略。第三部分考虑该分布式算法在复杂环境中的应用,将各种环境因素融入到算法的统一框架中,包括在噪声探测模型、扩展目标和多目标环境中重新定义跟踪质量函数,在障碍物环境中将障碍物作为虚拟邻节点加入连接势函数等。本文的主要贡献可分为四个方面。首先,首次提出将节点运动节能作为移动传感器网络跟踪的设计重点,而目前的移动传感器网络跟踪研究一般都只把跟踪性能作为唯一指标。其次,首次提出将跟踪性能和各种跟踪约束(包括节能、网络连通、避撞等)用目标函数量化表示,使跟踪问题可转化为经典的数学问题——多目标优化问题。第三,提出了一种完全分布式的跟踪算法,即每个节点仅依靠其邻节点信息决定自身控制输入,不存在任何动态中心节点,所有节点地位平等,提高了系统的灵活性和鲁棒性。第四,建立了一个全面的跟踪方案框架,适用于包括异构网络、噪声探测、障碍物、扩散目标、多目标等各种复杂环境,不需要特别针对某种应用环境开发新的跟踪算法,具有良好的可扩展性。更多还原

【Abstract】 This dissertation studies the problem of target tracking using mobile sensor networks.It focuses on using the capabilities of sensing,communication and locomotion of sensor nodes to keep tracking unknown mobile targets within the sensing area.A mobile sensor network is a wireless network comprising a large number of mobile sensor nodes.These nodes are always low-power and destructible.So energy saving and distributed computation are the most important factors in the tracking design.This dissertation proposes a fully distributed algorithm for target tracking under complicated environment such as noisy sensing,obstacle,diffused target,etc.It tries to maintain the targets being visible to the network all the time while consuming as little motion energy as possible.Meanwhile the network connectivity is maintained,node collision is avoided,etc.This dissertation work is composed of three parts.First,while minimizing the tracking energy consumption during the tracking process is proved to be NP-complete,an approximately optimal solution named breadth-first leader-follower algorithm is presented.It only can be realized in a locally centralized manner.We prove that its consumption is within a scalar factor of the optimal consumption.Second,we deal with the distributed tracking problem which is more practical in reality.The system modeling is set up and the data transmission is scheduled carefully so that only a few nodes are activated while other nodes keep idle.Then a motion strategy for active nodes is proposed which is made up of three modules:ⅰ) the tracking quality function which reflects the tracking accuracy,ⅱ) the potential function which reflects the network connectivity status,ⅲ) the distributed energy saving motion strategy.Third,this algorithm is extended to complicated environment.Various environmental factors are added to the uniform tracking framework.For example,we improve the tracking quality function for the noisy sensing model,diffused target and multi-target.Also the obstacle is considered as virtual neighboring nodes in the potential function.The major contribution of this dissertation lies in four aspects.First,it is the first work that focuses on the energy saving problem in target tracking,while most existing works using mobile network only limit their problems on improving the tracking accuracy.Second,certain objective functions are defined to quantify the main requirements and constraints such as the tracking quality,the network connectivity status and the collision avoidance.They transform the tracking into a classical multi-objective optimization problem.Third,the optimization algorithm is fully distributed.No centralized processor or dynamic central node is needed.Each node acts as the same role,which makes the system more robust and flexible.Fourth,a general tracking framework is set up.It can be applied to many complicated situations such as the heterogeneous network,noisy sensing,diffused target,obstacle,etc.We need not to design a special algorithm for certain application,which means our algorithm has good expansibility.更多还原