Methodology and architectures for system-on-chip real-time control

Control systems are being embedded in an ever-increasing number of applications, many of them unfeasible prior to the joined advances in control, solid state technology and microelectronics design. In the era of information, communication and technology, society continuously absorbs new technologies and quickly demands new tools, services and commodities that require challenging control techniques operated at high sample rates. However, the implementation of modern control systems is by no means straightforward and requires multidisciplinary knowledge in fields of control, hardware and software design. Complex control laws with demanding real-time constraints are usually implemented using high-end devices combined with some custom logic in project-dependent, hence non-reusable solutions. Such high-end devices are intended to be flexible enough to cover a wide range of applications and therefore their architectural complexity often leads to inefficient control solutions in terms of power consumption, area and even performance. Furthermore, in order to fulfil the real-time requirements, complex device-dependent arithmetic routines have to be usually hand crafted in assembly language as most modern compilers are unable to make use of the narrow data vector instructions that are especially tailored for signal processing applications, and sometimes, the required performance can only be reached by a number of these devices running in parallel. This thesis presents a comprehensive study of embedded real-time control implementations, bringing together hardware/software co-design, development tools and controller formulation aspects under a common framework that leads to a novel design methodology. [Continues.