Skip to main content Skip to main navigation

Publication

Utilizing Multi-Agent Deep Reinforcement Learning For Flexible Job Shop Scheduling Under Sustainable Viewpoints

Jens Popper; William Motsch; Alexander David; Teresa Petzsche; Martin Ruskowski (Hrsg.)
International Conference on Electrical, Computer, Communications and Mechatronics Engineering, located at 2021 International Conference on Electrical, Computer, Communications and Mechatronics Engineering, October 7-8, Belle Mare, Mauritius, IEEE, 2021.

Abstract

Current trends place great demands on the flexibility and sustainability of modern production facilities. The optimisation of these Flexible Job Shop Scheduling Problems (FJSSP) under multiple objective variables, such as the makespan or the consumed energy, is a great challenge for today’s planning systems due to the constantly changing constraints. In this paper, we present a method for multi-criteria dynamic planning of production facilities under both common and sustainable target variables, based on a Multi-Agent Reinforcement Learning (MARL) procedure. This is experimentally applied to a planning problem in a series of trials and compared with common methods. Finally, the results and further research questions are presented.

Projekte