| Systems-on-a-Chip (SoC) applications have been the focus of active research over the past decade. Evolving from simple prototype applications, novel more sophisticated SoC are currently being developed for a variety of applications. The SoC market is expanding rapidly to areas like biomedical, mechatronics, avionics and automotive systems. A new generation of SoC are the integrated microchemical systems; miniature chemical systems that carry out chemical reactions and separations, in precisely fabricated three dimensional microreactor configurations. The main focus of research in microchemical systems remains devoted to the microfabrication aspects. Nevertheless, parallel to the current fabrication research efforts, modeling and control of microchemical systems have gained considerable attention.; From a theoretical perspective, the system states in microchemical systems are functions of space and time, thus Distributed Parameter Systems (DPS) with combined distributed boundary sensing and actuation. In the first part of the dissertation (Chapters 2 and 3), we address the challenges of developing reduced-order dynamical models that capture efficiently the behavior of these SoC, and providing closed-loop boundary control schemes of reduced complexity. We propose using second-order statistical properties, to obtain a set of empirical eigenfunctions which are subsequently used to obtain finite dimensional low-order dynamical systems. Furthermore, we propose a novel boundary control scheme using empirical eigenfunctions in a constrained optimization procedure that can track a desired spatiotemporal profile.; From a practical perspective, we examine the possibility of integrating miniature systems with sensors, actuators and control electronics on a single chip to create a highly reliable microplant. In particular we investigate pathways for the implementation of Model Predictive Control (MPC) on-a-chip, focusing on microchemical but also drug-delivery systems. Considering the advantages of MPC, such as the ability to handle constraints, nonlinear processes and multivariable problems, it is of considerable importance to expand its applicability to dynamical systems which have fast response times-traditionally considered unsuitable for MPC. In the second part of the dissertation (Chapters 4-6), three pathways for the implementation of optimization-based control on-a-chip are presented: an application-specific instructor processor implementation, a general purpose processor implementation, and finally a codesign approach that leads to a field programmable gate array implementation. |
