Evolutionary fuzzy real-time job-shop scheduling

Real time task scheduling can be a challenging problem because of inherent system uncertainties such as task importance, timing and computation time, and particularly when the system is under overload, i.e. it is given more tasks than it can possibly complete in the allotted time span. To alleviate these problems, we first propose a novel fuzzy scheduling approach in which the real time scheduling is treated as a multi-criteria optimization problem. Consequently genetic algorithms are applied to optimize membership functions of the resulting fuzzy systems. Simulation results indicate that the proposed fuzzy scheduler increases both the total number of executed tasks as well as number important tasks that are completed, when compared with the bench mark approach Application of genetic algorithms to membership function optimization further improves these results.