Predictive-Delay Control for Overloading in Real-Time Scheduling

Abstract

In multi-task control systems, control quality is subject to deterioration due to system dynamics and several scheduling artefacts. Based on open-loop scheduler or feedback scheduling, we investigate the advantage of some new techniques, such as the subtask scheduling and the predictive delay control, used to deal with control and scheduling co-design constraints. In the latter, at each time instant, the measurement signal is predicted by extrapolation that minimizes the effect of the measurement obsolescence. This predictive method, compared to other complex dynamic methods, is easier to formulate and its results within a discrete-time control algorithm are suitable for embedded systems. In the present work, simulations are conducted to show that the predictive-delay control can improve the control quality even in the absence of a dynamic priority assignment like in the Earlies deadline First Algorithm. However, in order to take advantage from the potential of both methods namely the predictive-delay control and the subtask-scheduling, another alternative is to combine them in the same solution to better deal with the input-output latency. The experimental validation is accomplished using the servo-motor and the inverted-pendulum systems through a stochastic execution-time implementation.