基于能量调度的信息物理系统远程估计问题研究

发布时间：2019-07-01 09:39

【摘要】：随着信息技术的飞速发展,物理世界的各类设备逐渐互联互通,深度融合信息空间和物理空间的信息物理系统(Cyber-PhsicalSystems,CPS)应运而生。它集成计算、通信和控制技术,通过资源优化配置,感知、管理、调控着人类生活的方方面面,在工业生产、智能电网、智能交通、智慧医疗和国防军事等领域具有广泛应用,极大地改善了人类生产生活方式,因此受到国内外研究者的高度关注。信息物理系统的众多应用中都需要对系统状态进行远程估计,例如环境监测中对污染指标信息的准确估计决定着监测效果,工业生产中对被控对象运行状态的精确估计影响着控制信号的计算和生产效率。因此远程估计问题是CPS的基本问题之一,准确高效的远程估计是系统综合性能得以保障的先决条件。然而网络节点资源受限、通信不可靠和安全威胁给远程估计的研究带来了巨大挑战。最近,基于能量调度的CPS远程估计研究得到了广泛关注并取得了很大进展,但仍存在许多不足之处。例如,当前利用实时信息进行资源在线调度的研究缺少考虑时变通信环境的影响,对于攻击行为的认知主要局限于典型单一攻击模态以及远程估计的应用研究多数基于对物理对象和通信环境的理想假设等。本文针对这些研究不足,从远程估计的性能优化、安全保障和技术应用三个层面开展研究,主要工作和贡献点概括如下:1.概述了信息物理系统远程估计问题及其研究现状,并分析了已有研究的不足之处。2.研究面向远程估计的传感器能量最优动态调度策略。考虑具有时间关联特性的时变衰减信道对远程估计性能的影响,以最小化远端估计器期望平均估计误差协方差为优化目标,以传感器平均能量消耗为约束,利用Markov决策过程理论设计了一种基于实时丢包信息的动态能量调度策略,并从理论上证明了动态调度的最优性。对于传感器每时刻只能用低能量发送本地数据的情形,证明并给出了远程估计稳定性的充分必要条件。3.研究面向远程估计的随机事件驱动能量调度机制。为精确刻画事件驱动调度下远程估计的先验分布,引入具有时变参数的随机事件驱动机制,利用系统实时测量新息设计了更加有效的在线能量调度策略,得到相应的极小均方估计算法。提出了调度策略中最优化参数序列的一般性问题,并且对于一类具有特殊参数结构的应用场景,给出了最优的参数切换策略。4.研究针对远程估计的事件驱动攻击策略和性能评估。考虑了一种具有入侵传感器、篡改调度算法、窃听测量信息和注入错误反馈信息等多种能力的攻击者,设计了能够躲避攻击检测的事件驱动智能攻击策略。从攻击者角度获得了攻击下的近似极小均方估计算法,并解析地给出攻击过程中不改变通信率的攻击阈值。为评估估计性能,利用随机分析理论证明了所设计的攻击策略破坏了远程估计稳定性条件和渐近估计精度。5.研究基于远程估计的能量有效目标追踪方法。在感知距离受限移动传感器进行目标追踪的应用场景中,考虑不可靠通信导致数据丢包的影响,构建了保障传感器能够以预先给定概率追踪到移动目标的约束下最小化总移动距离的优化问题。基于远程估计技术和最优任务分配理论的应用,设计了包含目标位置估计与预测、目标不确定区域覆盖、移动节点分配三个步骤的能量有效协同追踪方法。最后对全文进行了总结,并展望进一步的研究工作。
[Abstract]:With the rapid development of information technology, the information physical systems (CPS) of the information space and physical space of the advanced integration information space and physical space have come into being. It integrates computing, communication and control technology, and has wide application in the fields of industrial production, intelligent power grid, intelligent transportation, intelligent medical treatment and national defense military through resource optimization, configuration, sensing, management and control of all aspects of human life. The invention greatly improves the human production and life style, and is therefore highly concerned by the researchers at home and abroad. In many applications of the information physics system, it is necessary to estimate the system state remotely, such as the accurate estimation of the pollution index information in the environment monitoring, and the accurate estimation of the running state of the controlled object in the industrial production affects the calculation and production efficiency of the control signal. Therefore, the remote estimation problem is one of the basic problems of CPS, and the accurate and efficient remote estimation is a prerequisite for guaranteeing the comprehensive performance of the system. However, the resources of the network node are limited, the communication is unreliable and the security threat poses a great challenge to the research of the remote estimation. Recently, the research of CPS remote estimation based on energy scheduling has received extensive attention and made great progress, but there are still many shortcomings. For example, the current research of resource on-line scheduling with real-time information lacks the influence of time-varying communication environment, and the cognition of the attack is mainly limited to the typical single attack mode and the application of the remote estimation, most of which are based on the ideal hypothesis of physical object and communication environment. In view of these deficiencies, this paper studies the performance optimization, security assurance and technical application of remote estimation, and the main work and contribution points are as follows:1. In this paper, the problem of remote estimation of the information physical system and its research status are summarized, and the shortcomings of the existing research are also analyzed. The energy-optimal dynamic scheduling strategy for remotely-estimated sensors is studied. considering the effect of a time-varying attenuation channel having a time-related characteristic on the performance of the remote estimation to minimize the expected average estimated error covariance of the remote estimator as an optimization target, the sensor average energy consumption is constrained, The dynamic energy scheduling strategy based on real-time packet loss information is designed by the Markov decision process theory, and the optimality of the dynamic scheduling is proved theoretically. For the case where the sensor can only transmit local data at low energy per time, a sufficient and sufficient condition for the stability of the remote estimation is given. The stochastic event-driven energy scheduling mechanism for remote estimation is studied. In order to accurately describe the prior distribution of the remote estimation under the event-driven scheduling, a random event-driven mechanism with time-varying parameters is introduced, and a more effective online energy scheduling strategy is designed by using the system real-time measurement of the new interest, and a corresponding minimum mean-square estimation algorithm is obtained. The general problem of the optimal parameter sequence in the scheduling strategy is put forward, and the optimal parameter switching strategy is given for a class of application scenarios with special parameter structure. The event-driven attack strategy and performance evaluation for remote estimation are studied. An event-driven intelligent attack strategy, which is capable of avoiding attack detection, is designed in this paper, which has the capability of intrusion sensor, tamper scheduling algorithm, eavesdropping measurement information and injection error feedback information. The approximate minimum mean-square estimation algorithm under the attack is obtained from the attacker's point of view, and the attack threshold of the communication rate is not changed in the attack process. In order to evaluate the estimation performance, the random analysis theory is used to prove that the designed attack strategy has destroyed the remote estimation stability condition and the asymptotic estimation precision. The energy-effective target tracking method based on remote estimation is studied. In the application scene of sensing distance limited movement sensor for target tracking, considering the influence of unreliable communication on the data packet loss, the optimization problem of minimizing the total moving distance under the constraint that the sensor can track the moving object with the preset probability is constructed. Based on the application of the theory of remote estimation and optimal task assignment, the energy-effective cooperative tracking method is designed, which includes three steps: target position estimation and prediction, target uncertainty area coverage, and mobile node assignment. Finally, the paper sums up the full text, and looks forward to the further research work.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP29

【相似文献】