Title: Q-learning and ACO hybridisation for real-time scheduling on heterogeneous distributed architectures

Authors: Younes Hajoui; Omar Bouattane; Mohamed Youssfi; El Houssein Illoussamen

Addresses: Laboratory SSDIA, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia 28999, Morocco ' Laboratory SSDIA, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia 28999, Morocco ' Laboratory SSDIA, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia 28999, Morocco ' Laboratory SSDIA, ENSET Mohammedia, Hassan II University of Casablanca, Mohammedia 28999, Morocco

Abstract: In the intensive computation field, it is worth mentioning that extensive computing power and considerable storage capacity are needed by greedy applications. To reach the required processing power, multiple processing units should be linked to handling the distributed jobs. However, the heterogeneity of the associated resources/workers is to be considered during the tasks scheduling process. Our approach consists of combining Q-learning with ant colony optimisation (ACO) to solve the job scheduling problems on heterogeneous architectures. The proposed approach is implemented by using the mobile agent's systems. The obtained results from simulation demonstrated the effectiveness of the proposed hybridisation due to the considerable reduction of the overall execution time (makespan) and to the fast convergence observed after a small number of learning steps.

Keywords: task scheduling; hybridisation; mobile agent system; Q-learning; ant colony; makespan.

DOI: 10.1504/IJCSE.2019.103826

International Journal of Computational Science and Engineering, 2019 Vol.20 No.2, pp.225 - 239

Received: 22 Mar 2018
Accepted: 09 Jan 2019

Published online: 27 Nov 2019 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article