Design of Automotive Control Systems Robust to Hardware Imprecision

Abstract

The hardware used for software implementation on a physical system introduces uncertainty into the controller. If neglected during design, this uncertainty can lead to poor controller performance, resulting in significant design and verification iterations. In this work, the effect of sampling time, quantization, and fixed-point computation are directly accounted for in the control design. Sampling time is compensated for by a discrete-time controller. A generic methodology is developed for modeling the worst-case scenario effect of quantization and fixed-point computation on the control commands. The cold-start emission control problem is used as a case study, and a discrete-time sliding surface controller is developed. Verification is performed to ensure the estimated worst-case scenario uncertainty bounds are accurate. The bounds are incorporated into a modified version of the control laws. During simulation the modified controller demonstrates significant reduction in tracking error in the presence of hardware imprecisions.