The integration of computers, communication, and control in modern distributed systems such as intelligent transit systems and industrial facilities has garnered a great deal of attention in recent years. This thesis presents an investigation into basic research, technology development, and implementation of monitoring and control strategies for networked-enabled dynamical systems. In view of the hierarchical nature of these systems, the thesis focuses on two important system layers;​ namely, networked control systems within the direct-control (servo) layer, and remote monitoring and supervisory control within the supervisory-layer.

In the present work, a networked control system (NCS) comprises the traditional components of a control system such as a plant, sensors, actuators, controllers, and signal conditioning and modification devices, but may be geographically distributed and linked through a communication network. In particular, in an NCS, the traditional feedback control loops are closed through a real-time communication network. A primary objective of the present thesis is to investigate, analyze, develop, implement and test stable control strategies to overcome the transmission problems between sensors, actuators, and controllers in an NCS;​ specifically, non-deterministic delay, losses, vacant sampling, and mis-synchronization of data/information. The term networked control is used in this thesis to mean the control of a networked system, and a networked system with a controller as an integral part is an NCS.

In this thesis, a new networked control strategy is developed which takes advantage of the potential capability of constrained Model Predictive Control (MPC) to compensate for anticipated data transmission problems. The constructive and computationally inexpensive strategy, which is developed here, uses predictions of future control actions and estimations of possibly erroneous sensory signals, to maintain good performance and stability of a closed-loop NCS, through an innovative method of information buffering. Network and computational loads are effectively minimized by incorporating variable prediction horizons. Performance is further improved through adaptive weight sequencing enhancements.

The analytical issues of system stability of the new networked control strategy are rigorously treated in the thesis. The analysis is carried out in two stages. In the first stage, prefect state measurement is assumed to develop the necessary and sufficient conditions for guaranteeing closed-loop asymptotic stability while utilizing predicted future control actions that are suboptimal. Stability is achieved in the sense of Lyapunov by imposing a terminal cost and bounding it without end constraints, and projecting the instantaneous states to the terminal states. Feasibility, which in turn implies stability, is proven in the context of this suboptimal control strategy. The stability surfaces generated from the resulting theorems are particularly useful as design guidelines to determine the effective worst-case delay that can be sustained by the networked control strategy. The second stage of stability analysis assumes imperfect state measurement with the purpose of determining the amount of deviation between the actual states and the estimated states that can be sustained in stable manner by the developed networked control strategy. This is accomplished by establishing bounds for the evolution of future inputs and states, employing the concept of perturbation sensitivity in nonlinear programming, and based on Lyapunov's second method. The theorems that result from the second analytical stage serve as a precise criterion in the design of extended state observers for compensating possible transmission problems between sensors and controllers of a networked system.

The functionality of a networked control system may be further enhanced by integrating the capability of remote monitoring and supervisory control into the upper layers of the hierarchy of the system. As a contribution in this direction, a framework for developing a universal and scalable network infrastructure for web-based monitoring and supervisory control of dynamic systems is developed in the present work. Practical implementation of this framework is detailed, using a low cost and flexible two-tier client-server architecture, where a user is able to remotely carry out a variety of tasks including system operation, experimentation, process monitoring and supervision, task scheduling, system reconfiguration and tuning, control, and safety/emergency routines, through a web-browser interface. A single web-server provides smooth information flow using a robust and intelligent scheduling scheme. The flexibility and modularity of the developed network architecture provide the rationale for further incorporation of a multi-layered intelligent supervisory control structure with the objective of on-line improvement of the performance of a process. A remote supervisor, which incorporates knowledge-based decision making, continuously monitors the performance of the process to infer the best adaptive controller (and the best tuning actions, if needed) for the process under the existing conditions.

The strategies of networked-enabled monitoring and feedback control, as developed in the present work have been implemented on an industrial fish-processing machine using an Ethernet communication network. The developed and analyzed technologies are thoroughly evaluated and validated through experimentation under a variety of operating conditions in real-time, and are demonstrated to be practical, beneficial, and technically sound.