The verification and validation step is a major milestone in the development of an industrial control system. Verification refers to showing that the control software is built correctly, and validation refers to showing that the controller meets its specified objectives. Current verification and validation procedures are costly and time consuming. For example, verification and validation of an automotive electronic control unit can take two man-years and cost 5-6 million dollars. However failure to find design or implementation errors in systems such as engine control, braking, or steering risks terrible economic and human consequences. Large-scale deployment of complex interacting systems, such as self-driving cars, further increases these risks. This project will show how verification and validation costs can be greatly reduced by bridging the gap between the disciplines of control design and software and hardware engineering. To do this, constraints such as data sampling rates, word length, and arithmetic processing are incorporated throughout the control design process. This Grant Opportunity for Academic Liaison with Industry (GOALI) project will show that by satisfying specified robustness properties during design, the controller can be verified and validated with a minimum of further effort. Verification and validation is of critical importance to the automotive industry, and engineers from Toyota Motor Engineering & Manufacturing North America are partners in this project. However the results will also benefit the aerospace, power generation, and manufacturing industries.
A large number of errors detected during verification and validation are introduced during the initial stages of controller development. A critical gap occurs when uncertainty in controller software/hardware implementation is not considered as part of the controller design. This gap leads to the need for many Verification and Validation iterations and results in costly controller design. This project intends to fill this gap by (i) modeling and quantification of uncertainties that arise from controller implementation imprecisions, (ii) design of robust controllers to overcome implementation uncertainty, and (iii) development of an adaptive control framework to update uncertainty bounds from implementation imprecisions. The expected outcome from these three main contributions will be an uncertainty-adaptive, easily verifiable control theory framework that industry can adapt to controller design processes to minimize the time and cost of controller development.
