Performance optimization for real-time scheduling systems based on fuzzy predictive control

This paper presented a new method for performance optimization of real time scheduling systems. The proposed method is based on fuzzy predictive control, in which the control objectives of system utilization and task deadline miss ration are formulated by their fuzzy membership function respectively, and the cost function of the predictive control is then represented as a fuzzy decision function combining two fuzzy objectives. The decision problem is solved using branch and bound technique, its result is an utilization increment which maximize the satisfaction level of the decision function. Then the proper task rates are computed by the task rate regulating algorithm. Simulation results and analysis show that the approach can achieve a good tradeoff between high system utilization and low tasks deadline miss ratio.