Design of trusted piecewise-affine controllers and virtual sensors into CMOS integrated circuits

Resumen

This Ph.D. Dissertation presents hardware realizations to generate PWA functions that are easy to configure and program, and that allow including embedded controllers and sensors into challenging applications requiring very small size, low power consumption, rapid response, short time-to-market, and security in the data provided by the hardware as well as integrity of the hardware itself. Digital architectures that overcome the proposals in the state-of-the-art are proposed to generate PWA functions. To implement and verify them, a hardware design flow is developed and employed with Application Specific Integrated Circuits (ASICs) and programmable devices as Field-Programmable Gate Arrays (FPGAs). In order to reduce time-to-market of these designs, an automated and integrated design flow is used to apply them as embedded controllers and virtual sensors. In the case of embedded control, the design flow is based on a model-based methodology, which applies model predictive control (MPC) and uses a Matlab toolbox to extract the PWA controllers from mathematical models of the physical system to control, obtaining the parameters needed to program and configure the proposed ASIC and FPGA realizations. Several cases of study are addressed to illustrate the highperformance in terms of area, power, and speed of the resulting embedded controllers into nanometer CMOS integrated circuits. In the case of embedded sensors, PWA virtual sensors are employed to model nonlinear relations between input variables which are easy to measure and output variables which are costly to evaluate directly. The parameters needed to program and configure the proposed ASIC and FPGA realizations are obtained with an identification algorithm that uses numerical input-output data coming from experiments or simulations. Since PWA systems are a subset of fuzzy systems, a design methodology based on Xfuzzy environment is presented. A case of study from the automotive domain is selected to illustrate the performance of the resulting nanometer CMOS integrated circuits. The inclusion of security is crucial in embedded control and sensing systems to guarantee not only data privacy and integrity but also device authenticity. Hence, another challenge addressed in this Ph.D. Dissertation is the design of trusted PWA hardware. For that purpose, lightweight cryptographic primitives are analyzed and the most suitable are selected to be included in the proposed ASIC realizations. A 90-nm CMOS ASIC prototype of trusted virtual sensor that offers privacy, authenticity and integrity of the measurements and the circuit itself is presented.