UC Berkeley

The increasing complexity in automotive systems has prompted the use of the Verification & Validation procedure in the industry. In particular, powertrain control design follows the validation of achieving the control targets and the verification that the correct controllers were selected. Within this process, there is an area of significance residing in the early stages of Verification & Validation that addresses issues arising when digital controllers are implemented on continuous time physical systems. The work in this thesis explores that realm via the design of control algorithms that mitigate errors and uncertainties in the software-hardware interface. In particular, control algorithms that mitigate sampling time, quantization, fixed-point processing, and model uncertainty are presented. The formulations include discrete time sliding control, discrete time sliding control with uncertainty bounds, and decoupled as well as multiple-input, multiple-output discrete time sliding control with adaptation. The algorithms are demonstrated on engine cold start dynamics, with a secondary objective of addressing the issue of hydrocarbon emission reduction, with simulation and experimental results.