【作者】 任祝；

【导师】 沈学民； 孙优贤； 陈积明；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2013， 博士

【摘要】 随着微机电系统(MEMS)与无线通信技术的发展,小型、廉价的无线传感器渐渐进入人们的视野。由于无线传感器能携带声学、光学、视频与无线通信等模块,且具有感知、处理和收发数据等多种功能,所以当它们被大量地分布时,组成的无线传感器网络可以联合监视、控制某区域。因此,无线传感器网络在安全防卫、控制系统、医疗等各领域具有广泛应用前景,针对它的研究也越来越受到关注。通常情况下,传感器①使用电池或者外界可充电供电。由于昂贵的人工费,环境与网络规模等因素,传感器经常被要求在无人维护下正常工作较久的时间。能量问题对传感器的工作、通讯次数与布置都造成了很大的限制。为了节约能量,传感器通常会按照某种调度策略,在工作与休眠状态之间切换。由于状态的切换会中断传感器的运作,例如感知停止,降低对系统的估计性能等,为了提高能量利用率,许多致力于对传感器调度研究的工作已经大量开展。基于最近的研究成果,本文主要探索能量受限传感器调度问题,其中主要工作与贡献包括以下几个方面：1.概述了无线网络化控制系统及传感器调度问题的相关研究与进展。2.研究了可充电传感器网络的随机事件探测问题。在有限随机充入电量约束下,首先针对单传感器完全信息场景,设计了一种利用事件发生信息的最优调度策略,而且证明了它具有简单的结构。然后在不完全信息场景,提出了一种易于实现的启发式算法。最后将结果推广到多传感器场景,设计了令传感器相互合作的调度策略,优化网络的总体探测率。3.考虑了平均发送能量有限的情况下,优化远程状态估计器估计性能的周期调度问题。由于无线信道具有随机丢包性,在每个时刻传感器需要决策是否发送数据。通过马尔科夫链,得到了使估计器的平均误差协方差最小的最优的周期调度策略,分析了系统参数、能量约束、丢包率三个因素对估计器稳定性的影响。4.假设传感器有两个能量发送水平,在高能量水平下,数据传送具有相对较高的成功接收率,低能量水平下,具有较低的成功接收率。通过利用远程估计器的反馈信息,得到了结构简单、最优的动态调度策略,使得估计器的平均误差协方差在给定的平均能量限制下达到最小,而且得到了在最优动态调度策略下,使估计器稳定的充分必要条件。5.考虑了网络化控制系统中的调度问题,其中两个传感器分别观测不同的系统,然后将估计值通过网络发送到接收端处进行深层处理。由于传感器与中继节点的能量有限,它们需要合作地进行工作,优化全局能量利用率。本文通过将估计问题转化为等价的路由问题,提出了一种协作的网络调度策略使得两个传感器的估计性能同时得到优化。最后,对全文进行总结并讨论了进一步研究方向。更多还原

【Abstract】 With the rapid progress of Micro-Electro-Mechanical Systems (MEMS) and wireless communication technologies, small and inexpensive wireless sensors are getting possible. A large number of wireless sensors, which are capable of sensing, collecting and processing, can be deployed extensively to construct a wireless sensor networks and cooperate with each other to monitor a region of interest. Since the sensors will have acoustics, optics, video and wireless communication modes, they can collect useful information on entities of interest and send back the information to base station. Thus wireless sensor networks are widely used recently, especially in surveillance, control system, health care and agriculture etc.Generally, the sensors②are operated on battery power or slow environmental recharge process. Due to expensive manual maintenance, environmental and network size concerns, the sensors are required to work for a long unattended time. This places a hard, stringent energy constraint on the design of the operation, communication times and the deployment of these sensors. Because of this energy constraint, the sensors will be scheduled to turned on and off alternatively for saving energy. However, this alternation will inevitably dis-rupt sensors’ operation, e.g., stop sensing, weaken control of the system. Thus, in order to improve the energy efficiency, a lot of work has been devoted to sensor scheduling inves-tigation. Based on the latest results, this dissertation studies some specific problems for networked sensing and estimation. The main work and contributions are summarized as follows.1. A brief literature review on the development of wireless networked control systems, sensor scheduling and related works are provided.2. Two practical sensor scheduling for stochastic event detection are designed. The first one is for the single-sensor case, where the designed scheme will utilize the event’s information to deicide the sensor’s switching between active and inactive states. And then extend the results to the multiple-sensor case and a sensor schedule which will commands the sensors to detect the events cooperatively is proposed.3. The periodic sensor scheduling for remote state estimation under average transmis-sion energy constraint is considered. The sensor decides whether or not to send its data to a remote estimator due to the imperfect communication. An optimal periodic schedule is found via the tools from Markov chain. Furthermore, a sufficient condi-tion of the system dynamics, energy budget and packet drop rate, under which the remote estimator is guaranteed to be stable, is derived.4. Assume that the sensor has two transmission energy levels, where the high level cor-responds to a high packet reception ratio. By exploiting the feedback information from the remote estimator, a simple and optimal dynamic schedule which minimizes the average estimation error under the energy constraint is developed. The necessary and sufficient condition under which the remote state estimator is stable is also found.5. A schedule problem for a networked control system is considered, where two sen-sors will observe different systems, respectively, and then send the estimation data to a sink node through a network for further processing. Since the sensors and re-lay nodes’ energy are limited and they must work cooperatively, by transferring the estimation problem to an equivalent routing problem, a networked and cooperative sensor scheduling is proposed.In the end, the dissertation is concluded and some future research work are discussed.更多还原