This research project deals with the analysis, design,and implementation of robust controllers for nonlinear systems such as operating under variable conditions such as internal combustion IC engines. The operation of modern IC engines for automotive use must satisfy a diverse range on conflicting constraints imposed by economical, environmental , societal and marketing factors, such as; regulating exhaust emissions to near zero levels to meet government standards, providing increased fuel economy consistent with consumer acceptance, maintaining satisfactory vehicle performance over a wide range of operating conditions, and performing safely and reliably over the life of the vehicle. At the present time there exists no unified control methodology to provide robust controllers for nonlinear systems with time-delays, saturation constraints and large operating parameter variations. In addition, since modern controllers are to be implemented in a digital microprocessor, issues related to the implementability of these controllers are essential, but are often ignored. Hardware limitations and real-time computational constraints require the implementation of low-order digital controllers. This research project addresses the important and challenging short comings and limitations of modern control theory. Novel tools based on recently developed robust gain-scheduled control, fixed-order control sampled-data systems, alternating projection algorithms and linear matrix inequality methods are developed. The model of a diesel engine, provided by Cummins Engine Company, is used as a design paradigm to demonstrate the applicability of the developed techniques. Cummins has agreed to collaborate closely with the investigators by providing industrial expertise, simulation models, test data, and access to their experimental test facilities. The research results provide valuable insight and optimized design methodologies to control a large class of practical control engineering problems,such as engines, robots and manufacturing processes in a unified and computationally efficient manner.
