A Robust Control Real‐Time Scheduling Design

Abstract

AbstractToday's real‐time systems need to be operated under tighter performance specifications, and require more and more constraints to be satisfied. These specifications can only be met when system nonlinearities and constraints are explicitly considered. The paper presents a new H∞‐nonlinear model predictive control scheduling method to regulate the deadline miss ratio and CPU utilization, and to satisfy the performance specifications. It is difficult to accurately model a real‐time multimedia system. H∞ control theory addresses the issue of worst‐case controller design for linear plants subjects to unknown additive disturbances and plant uncertainties. Model predictive methodology is extended to H∞ controllers for nonlinear systems. A scheduling architecture that includes model uncertainty is proposed. We firstly integrate H∞ robust optimal control and scheduling theory in order to satisfy trade‐offs between control performance and computing resource utilization. Performance evaluation results are demonstrated by simulated examples, and earliest‐deadline first (EDF) scheduling results are compared.