A Multi-objective Task Scheduling Method for Embedded System Design

摘要

Efficient embedded system design requires considering several design parameters during the task scheduling step. In this paper, a new multi-objective task scheduling method based on genetic algorithm is proposed for embedded systems. In this method, the architecture platform and the tasks in the form of task graphs are given as the inputs of the algorithm. The objective functions in the proposed multi-objective task scheduling include reliability in addition to execution time and energy consumption. The experimental results show that, the proposed algorithm provides better solutions (i.e. scheduled tasks) in terms of all objectives. Moreover, in order to verify the optimization provided by the proposed algorithm, it is shown that the algorithm achieves better solutions in terms of each objective when compared to the solutions obtained by the greedy method. Furthermore, the efficacy of the proposed method is shown in comparison to some well-known single objective heuristic scheduling algorithms where the performance of the proposed method is 29.5% and 21% higher in terms of metrics of scheduling length ratio (SLR) and speeds up, respectively.