Tracking Control for Nonlinear Systems Distributed over Communication Networks

现代工程系统非常复杂，由大量相互连接的子部件组成，由于通信和电子技术的显著发展，这些子部件可以分布在广泛的区域并通过数据网络连接起来。对这些复杂系统的控制只能以分散的方式实现，方法是对每个部件或一小组部件进行适当的局部控制器设计。控制和计算能力分布在系统中，组件之间需要稳定的数据交换，才能使系统正常运行。虽然控制、信息论和通信在某种程度上是成熟的学科，但到目前为止，关于信息论和通信约束中的问题如何影响分布在通信网络上的大规模控制系统的性能的研究还很少。这种情况最简单的情况是，当控制功能由位于受控对象相同物理位置的本地(分散)控制器和位于不同位置的远程(集中式)控制器执行时。在这种情况下，典型的情况是要求一个子系统执行取决于驻留在远程物理位置的系统的状态的任务。在这种情况下，本地控制器必须嵌入一些要执行的任务的预测机制，这取决于远程子系统的状态，可能是相当不确定的。成功处理大不确定性的能力是自动控制的主要原因之一，如果不是最重要的一个原因的话。然而，到目前为止，很少有人注意到这样一个事实，即不确定性也可能影响控制系统所要求完成的任务。提出的研究活动的主要智力目的之一是展示如何通过在控制器中嵌入要执行的任务可能依赖的系统的可调内部模型来成功地解决控制任务不确定的问题。另一个主要的学术目的是展示这种设计理念如何特别适合于控制分布在通信网络上的大型复杂系统。事实上，内部模型的作用是实时生成执行实际控制任务所需的精确前馈输入。如果本地控制器包含这些任务的内部模型，远程控制器的作用将只是根据被调节变量的实际值重新设置(偶尔，可能使用少量信息)内部模型的状态。这将提供一种设计哲学，它将更好地处理分布在通信网络上的控制系统的典型限制，例如带宽限制、延迟、分组丢失。基于内部模型的控制设计的进展预计将对控制技术的各个领域产生广泛的影响。例如，实验结果已经表明，自整定内模控制器如何自动补偿旋转电机中发生的故障，以及这一具体应用可能给制造业带来的好处是非常明显的。另一个令人感兴趣的广泛领域是电信卫星控制。在这个领域发生的典型问题是通信卫星在圆形轨道上以固定距离跟踪较小的信标卫星，或者小卫星网络处于协调运动中。另一个初步模拟结果令人印象深刻的应用是设计一种自动驾驶仪，帮助救援直升机降落在公海上遇险的船只上。类似的应用包括在恶劣大气条件下自动装船/卸货，或遥控潜水艇进行自动潜水或管道检查。

Modern engineering systems are very complex and comprise a high number of interconnectedsub-components which, thanks to the remarkable development of communications and electronics,can be spread over broad areas and linked through data networks. The control of those complex sys-tems can only be achieved in a decentralized mode, by appropriately designing local controllers foreach individual component or small group of components. Control and computational capabilitiesbeing distributed over the system, a steady exchange of data among the components is required, inorder for the system to behave properly. While control, information theory and communication areto some extent mature disciplines, little effort has been put so far in understanding how issues ininformation theory and communication constraints affect the performance of a large-scale controlsystem distributed over a communication network. The simplest case in which this situation mayoccur is when the control functions are performed by a local (decentralized) controller, situated atthe same physical location of the controlled plant, and by a remote (centralized) controller, situatedat a different location.Typical, in this scenario, is the situation in which a subsystem is required to perform a taskwhich depends on the state of a system residing at a remote physical location. In this case, thelocal controller must embed some mechanism of prediction of the tasks to be performed which,depending on the states of remote subsystems, may be quite uncertain. The ability of successfullyhandle large uncertainties is one of the main, if not the single most important, reasons of being ofautomatic control. Little attention, though, has been paid so far to the fact that uncertainties mayalso affect the task that the control system is asked to accomplish. One of the main intellectualpurposes of the proposed research activity is to show how problems in which control tasks areuncertain can be successfully addresses by embedding, into the controller itself, a tunable internalmodel of the systems on which the tasks to be performed may depend. The other main intellectualpurpose is to show how this design philosophy is particularly suited to the control of large complexsystems distributed over communication networks. In fact, the role of an internal model is thatof generating, in real time, the precise feed-forward input which would be required to perform theactual control task. If the local controller contains an internal model of those tasks, the role of theremote controller would only be that of resetting (once in a while, possibly using little amount ofinformation) the state of the internal model on the basis of the actual value of the regulated variable.This would provide a design philosophy which would much better cope with restrictions typical ofa control system distributed over a communication network, such as bandwidth constraints, delays,loss of packets.Progresses in internal-model-based control design are expected to have a broad impact in var-ious areas of control technology. Experimental results have already shown, for instance, how selfa tuned internal-model-based controller can automatically compensate for faults occurring in a ro-tating electrical machine, and the benefits which may derive to manufacturing industry from thisspecific application are quite clear. Another broad area of interest is that of satellite control fortelecommunications. Typical problems occurring in this area are to have a communication satellitetracking a smaller, beacon satellite at a fixed distance in a circular orbit, or to have networks ofsmall satellites ying in coordinated motion. Another application, in which preliminary simulationshave shown impressive results, is the design of an autopilot to assist the landing of a rescue heli-copter on a ship in distress in high seas. Similar applications are automatic ship loading/unloadingin adverse atmospheric conditions or remote control of submarine vehicles for automatic underseacable or pipeline inspection.