实现网络控制系统自治化三大关键技术的算法研究

Nowadays, computer technology allows electronic equipment to surpass humans in decision making capabilities. Electronic equipment also has the advantages of being more versatile in many hostile environments that humans are unable to operate in. Because of these advantages, there is a growing interest in developing networked control systems to replace humans in these conditions. Our project concerns over three integrated topics that bring networked control systems to the next level of autonomy. It involves model checking and modeling of modular systems, autonomous information acquisition, and autonomous inter-controller communications. For modeling and model checking of modular systems, we design systems that are instantiated from templates, which allows rapid reproduction of the system components while enables an appropriate system reaction and can even maintain many functions when some of its components malfunction. For model checking, our approach is to refine the state space of templates in order to capture the interactions among components, which avoids traditional system synchronization and its fatal consequence: combinatorial explosion. Combinatorial explosion renders a system to be impractical to complete decision making processes. To research information acquisition, we let controllers decide what to observe, while taking the software and hardware constraints of practical systems into consideration. Our goal is to selectively gather only the information that is vital to control and decision making. This prevents data overload, a common cause of decision errors. In military applications, this can be vital in safeguarding valuable assets such as radar, because emitted radar signals poses a risk of being detected. To address the classical problem of inter-controller communications, we make controllers communicate with their system estimates instead of the event occurrences, which make us limit our previous algorithm for computing communication policies to acyclic systems. To test these results in real time, we'll conduct numerical simulations and hardware-in-the-loop experiments for closely related applications, including modeling and autonomous routing of unmanned aircraft systems, sensor scheduling for wireless sensor networks in industrial process control, and optimizing material handling systems in airports and ship harbors.

电子技术的迅猛发展使得电脑的决策水平于不久的将来超过人类，这使得网络控制系统自治化技术的发展成为电子信息领域竞争的焦点。本项目的研究对象是网络控制系统自治化的三大关键技术: 模板化模型的检测与建模、自治化信息采集、和控制器之间的自治化通讯。对于模板化模型的检测与建模，本项目将通过细化模板的状态空间来体现构件间的耦合关系，从而避免传统的系统整合及其导致的维度灾难问题。对于自治化信息采集，在控制器自主决定信息采集内容的基础上，我们将根据实际系统的物理限制来提出具备特殊应用背景的算法。对于如何优化控制器之间通讯这个经典难题，我们的重点是突破先前提出的高效算法对系统拓扑结构的要求，从而提出适合循环系统的算法。为了将理论应用到实践中去，我们将对相关应用题目来进行仿真模拟试验。这些应用具体包括无人侦察机系统模块化建模与自治化路由、工业流程控制中的无线传感网络传感器自治化调度、和机场码头物流控制。

网络控制系统自制化技术的是电子信息领域竞争的焦点之一。本项目针对网络控制系统自制化技术进行了广泛深入的研究，主要成果集中于以下几个方面：.关于于自制化信息采集的研究主要集中于满足监控目的的在线算法研究和满足故障诊断的离线算法研究。在研究在线算法中，团队成员延续了先前将监控对系统观测的要求转化成识别一些状态对中的状态的做法，在构建需要识别的状态对时将系统未来可能的变化纳入考虑，从而发展了仅需检查当前状态估计的高效在线感应器控制算法。对于自制化信息采集的离线算法主要集中在以系统故障诊断为目的的研究上，主要的突破是发展了一种没有额外解空间限制的算法。另外，针对监控的自制化信息采集，团队成员还发展了计算所有局部优化解的算法。这些成果共总结成三篇控制领域的权威杂志论文，两篇IEEE Transactions on Automatic Control论文和一篇Automatica论文。所有这些文章已经得到发表或被正式接受。.另外，本项目还对一些局部可观系统的性质也作了深入的研究。主要成果包括将相对可观性的问题转换成可观性的问题来解决。相对可观性和可观性都是针对系统监控的观察要求。业内认为可观性是相对可观性的一个特例。所提的转换方法不仅表明这两种性质在某种意义上是等价的，而且为使用解决可观性的方法来处理相对可观性的相关问题开辟了道路。 相关成果已经在控制界的权威学术会议IEEE Conference on Decision and Control 得到展示，相关的杂志论文投递到IEEE Transactions on Automatic Control。　.关于模型检测的研究主要集中在检测由相同构件组成但构件数目不定的系统。针对这种系统，团队成员发展了一种系统状态空间不随构建数目增加而变化的建模方法。然后，团队成员还发展了相应的模型检验的方法。这种方法的特点是能够直接计算出多少构件数目系统还不致于产生死结和活结。 同样，这种模型检验的计算复杂度并不随构建的数目变化而变化。另外，针对这种系统，团队成员还发展了一种用监控来消除死结和活结的方法。这些成果共总结成两篇学术论文，一篇被IEEE Transactions on Automatic Control接受，另一篇被Automatica接受。