Solving Decentralized Multi-Agent Pickup and Delivery Problems using Reinforcement Learning in a Simulated Environment

Keywords

Loading...
Thumbnail Image

Issue Date

2021-07-10

Language

en

Document type

Journal Title

Journal ISSN

Volume Title

Publisher

Title

ISSN

Volume

Issue

Startpage

Endpage

DOI

Abstract

Due to technological advancements the possibilities for robotic automa- tion are growing. There is increasing demand and necessity for robotic systems which intelligently and e ciently transport goods. Systems exist which can carry out such tasks, but they lack general employability, only being suitable for controlled environments at limited scale. Decentralized systems could solve these issues and allow robotic logistical systems to be more versatile and work on a larger scale. A simulation was created to test if reinforcement learning could be used to train agents in order to create such a decentralized system. It was found that in an abstract simulated environment this was possible. Furthermore, the results suggested that if robots in such a system can communicate information about their status to one another, the system's performance can be increased.

Description

Citation

Faculty

Faculteit der Sociale Wetenschappen