A multi-agent deep reinforcement learning framework for automated driving on highways

Louis Bakker; Sergio Grammatico

doi:10.1109/MED48518.2020.9182882

A multi-agent deep reinforcement learning framework for automated driving on highways

Team Bart De Schutter

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

2 Citations (Scopus)

Abstract

We apply deep reinforcement learning to automated driving on highways. We propose a novel, simple framework with improved performance with respect to the state of the art. When implementing our algorithm on multilane highway scenarios, after the training phase, we observe via numerical simulations that the vehicles are able to avoid collisions and to reach their respective destination lanes with very high probability.

Original language	English
Title of host publication	Proceedings of the 28th Mediterranean Conference on Control and Automation, MED 2020
Place of Publication	Piscataway, NJ, USA
Publisher	IEEE
Pages	770-775
ISBN (Electronic)	978-1-7281-5742-9
DOIs	https://doi.org/10.1109/MED48518.2020.9182882
Publication status	Published - 2020
Event	28th Mediterranean Conference on Control and Automation, MED 2020 - Saint-Raphael, France Duration: 15 Sept 2020 → 18 Sept 2020

Conference

Conference	28th Mediterranean Conference on Control and Automation, MED 2020
Country/Territory	France
City	Saint-Raphael
Period	15/09/20 → 18/09/20

Access to Document

10.1109/MED48518.2020.9182882

Cite this

@inproceedings{2b421b10622a44199128140c6ba6c4fe,

title = "A multi-agent deep reinforcement learning framework for automated driving on highways",

abstract = "We apply deep reinforcement learning to automated driving on highways. We propose a novel, simple framework with improved performance with respect to the state of the art. When implementing our algorithm on multilane highway scenarios, after the training phase, we observe via numerical simulations that the vehicles are able to avoid collisions and to reach their respective destination lanes with very high probability.",

author = "Louis Bakker and Sergio Grammatico",

year = "2020",

doi = "10.1109/MED48518.2020.9182882",

language = "English",

pages = "770--775",

booktitle = "Proceedings of the 28th Mediterranean Conference on Control and Automation, MED 2020",

publisher = "IEEE",

address = "United States",

note = "28th Mediterranean Conference on Control and Automation, MED 2020 ; Conference date: 15-09-2020 Through 18-09-2020",

}

Bakker, L & Grammatico, S 2020, A multi-agent deep reinforcement learning framework for automated driving on highways. in Proceedings of the 28th Mediterranean Conference on Control and Automation, MED 2020. IEEE, Piscataway, NJ, USA, pp. 770-775, 28th Mediterranean Conference on Control and Automation, MED 2020, Saint-Raphael, France, 15/09/20. https://doi.org/10.1109/MED48518.2020.9182882

A multi-agent deep reinforcement learning framework for automated driving on highways. / Bakker, Louis; Grammatico, Sergio.
Proceedings of the 28th Mediterranean Conference on Control and Automation, MED 2020. Piscataway, NJ, USA: IEEE, 2020. p. 770-775.

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

TY - GEN

T1 - A multi-agent deep reinforcement learning framework for automated driving on highways

AU - Bakker, Louis

AU - Grammatico, Sergio

PY - 2020

Y1 - 2020

N2 - We apply deep reinforcement learning to automated driving on highways. We propose a novel, simple framework with improved performance with respect to the state of the art. When implementing our algorithm on multilane highway scenarios, after the training phase, we observe via numerical simulations that the vehicles are able to avoid collisions and to reach their respective destination lanes with very high probability.

AB - We apply deep reinforcement learning to automated driving on highways. We propose a novel, simple framework with improved performance with respect to the state of the art. When implementing our algorithm on multilane highway scenarios, after the training phase, we observe via numerical simulations that the vehicles are able to avoid collisions and to reach their respective destination lanes with very high probability.

UR - http://www.scopus.com/inward/record.url?scp=85092180986&partnerID=8YFLogxK

U2 - 10.1109/MED48518.2020.9182882

DO - 10.1109/MED48518.2020.9182882

M3 - Conference contribution

AN - SCOPUS:85092180986

SP - 770

EP - 775

BT - Proceedings of the 28th Mediterranean Conference on Control and Automation, MED 2020

PB - IEEE

CY - Piscataway, NJ, USA

T2 - 28th Mediterranean Conference on Control and Automation, MED 2020

Y2 - 15 September 2020 through 18 September 2020

ER -

A multi-agent deep reinforcement learning framework for automated driving on highways

Abstract

Conference

Access to Document

Other files and links

Fingerprint

Cite this