Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Multi-agent Exploration with Reinforcement Learning
Department of Electrical and Computer Engineering, University of Patras, Greece.
Department of Electrical and Computer Engineering, University of Patras, Greece.
Luleå University of Technology, Department of Computer Science, Electrical and Space Engineering, Signals and Systems.ORCID iD: 0000-0003-0126-1897
Department of Electrical and Computer Engineering, University of Patras, Greece.
2022 (English)In: 2022 30th Mediterranean Conference on Control and Automation (MED), IEEE, 2022, p. 630-635Conference paper, Published paper (Refereed)
Abstract [en]

Modern robots are used in many exploration, search and rescue applications nowadays. They are essentially coordinated by human operators and collaborate with inspection or rescue teams. Over time, robots (agents) have become more sophisticated with more autonomy, operating in complex environments. Therefore, the purpose of this paper is to present an approach for autonomous multi-agent coordination for exploring and covering unknown environments. The method we suggest combines reinforcement learning with multiple neural networks (Deep Learning) to plan the path for each agent separately and achieve collaborative behavior amongst them. Specifically, we have applied two recent techniques, namely the target neural network and the prioritized experience replay, which have been proven to stabilize and accelerate the training process. Agents should also avoid obstacles (walls, objects, etc.) throughout the exploration without prior information/knowledge about the environment; thus we use only local information available at any time instant to make the decision of each agent. Furthermore, two neural networks are used for generating actions, accompanied by an extra neural network with a switching logic that chooses one of them. The exploration of the unknown environment is conducted in a two-dimensional model (2D) using multiple agents for various maps, ranging from small to large size. Finally, the efficiency of the exploration is investigated for a different number of agents and various types of neural networks.

Place, publisher, year, edition, pages
IEEE, 2022. p. 630-635
Series
Mediterranean Conference on Control and Automation (MED), ISSN 2325-369X, E-ISSN 2473-3504
National Category
Robotics
Research subject
Robotics and Artificial Intelligence
Identifiers
URN: urn:nbn:se:ltu:diva-92639DOI: 10.1109/MED54222.2022.9837168ISI: 000854013700103Scopus ID: 2-s2.0-85136300911OAI: oai:DiVA.org:ltu-92639DiVA, id: diva2:1689466
Conference
30th Mediterranean Conference on Control and Automation (MED), Vouliagmeni, Greece, June 28 - July 1, 2022
Note

ISBN för värdpublikation: 978-1-6654-0673-4 (electronic), 978-1-6654-0674-1 (print)

Available from: 2022-08-23 Created: 2022-08-23 Last updated: 2023-05-17Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Nikolakopoulos, George

Search in DiVA

By author/editor
Nikolakopoulos, George
By organisation
Signals and Systems
Robotics

Search outside of DiVA

GoogleGoogle Scholar

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 62 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf