Benefits of Social Learning in Physical Robots

Jacqueline Heinerman, Bart Bussmann, Rick Groenendijk, Emile Van Krieken, Jesper Slik, Alessandro Tezza, Evert Haasdijk, A. E. Eiben

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

Robot-to-robot learning, a specific case of social learning in robotics, enables the ability to transfer robot controllers directly from one robot to another. Previous studies showed that the exchange of controller information can increase learning speed and performance. However, most of these studies have been performed in simulation, where robots are identical. Therefore, the results do not necessarily transfer to a real environment, where each robot is unique per definition due to the random differences in hardware. In this paper, we investigate the effect of exchanging controller information, on top of individual learning, in a group of Thymio II robots for two tasks: obstacle avoidance and foraging. The controllers of the robots are neural networks that evolve using a modified version of the state-of-the-art NEAT algorithm, called cNEAT, which allows the conversion of innovations numbers from other robots. This paper shows that robot-to-robot learning seems to at least parallelise the search, reducing wall clock time. Additionally, controllers are less complex, resulting in a smaller search space.

Original languageEnglish
Title of host publicationProceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018
EditorsSuresh Sundaram
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages851-858
Number of pages8
ISBN (Electronic)9781538692769
DOIs
Publication statusPublished - 28 Jan 2019
Event8th IEEE Symposium Series on Computational Intelligence, SSCI 2018 - Bangalore, India
Duration: 18 Nov 201821 Nov 2018

Conference

Conference8th IEEE Symposium Series on Computational Intelligence, SSCI 2018
CountryIndia
CityBangalore
Period18/11/1821/11/18

Fingerprint

Social Learning
Robot
Robots
Controllers
Robot learning
Controller
Collision avoidance
Obstacle Avoidance
Foraging
Clocks
Robotics
Innovation
Search Space
Neural networks
Hardware

Keywords

  • Evolutionary Algorithms
  • Evolutionary Robotics
  • Neural Networks
  • Robot-to-Robot Learning

Cite this

Heinerman, J., Bussmann, B., Groenendijk, R., Krieken, E. V., Slik, J., Tezza, A., ... Eiben, A. E. (2019). Benefits of Social Learning in Physical Robots. In S. Sundaram (Ed.), Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018 (pp. 851-858). [8628857] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SSCI.2018.8628857
Heinerman, Jacqueline ; Bussmann, Bart ; Groenendijk, Rick ; Krieken, Emile Van ; Slik, Jesper ; Tezza, Alessandro ; Haasdijk, Evert ; Eiben, A. E. / Benefits of Social Learning in Physical Robots. Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018. editor / Suresh Sundaram. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 851-858
@inproceedings{c35f6d33a88a438988f1b06fa4151d6f,
title = "Benefits of Social Learning in Physical Robots",
abstract = "Robot-to-robot learning, a specific case of social learning in robotics, enables the ability to transfer robot controllers directly from one robot to another. Previous studies showed that the exchange of controller information can increase learning speed and performance. However, most of these studies have been performed in simulation, where robots are identical. Therefore, the results do not necessarily transfer to a real environment, where each robot is unique per definition due to the random differences in hardware. In this paper, we investigate the effect of exchanging controller information, on top of individual learning, in a group of Thymio II robots for two tasks: obstacle avoidance and foraging. The controllers of the robots are neural networks that evolve using a modified version of the state-of-the-art NEAT algorithm, called cNEAT, which allows the conversion of innovations numbers from other robots. This paper shows that robot-to-robot learning seems to at least parallelise the search, reducing wall clock time. Additionally, controllers are less complex, resulting in a smaller search space.",
keywords = "Evolutionary Algorithms, Evolutionary Robotics, Neural Networks, Robot-to-Robot Learning",
author = "Jacqueline Heinerman and Bart Bussmann and Rick Groenendijk and Krieken, {Emile Van} and Jesper Slik and Alessandro Tezza and Evert Haasdijk and Eiben, {A. E.}",
year = "2019",
month = "1",
day = "28",
doi = "10.1109/SSCI.2018.8628857",
language = "English",
pages = "851--858",
editor = "Suresh Sundaram",
booktitle = "Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

Heinerman, J, Bussmann, B, Groenendijk, R, Krieken, EV, Slik, J, Tezza, A, Haasdijk, E & Eiben, AE 2019, Benefits of Social Learning in Physical Robots. in S Sundaram (ed.), Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018., 8628857, Institute of Electrical and Electronics Engineers Inc., pp. 851-858, 8th IEEE Symposium Series on Computational Intelligence, SSCI 2018, Bangalore, India, 18/11/18. https://doi.org/10.1109/SSCI.2018.8628857

Benefits of Social Learning in Physical Robots. / Heinerman, Jacqueline; Bussmann, Bart; Groenendijk, Rick; Krieken, Emile Van; Slik, Jesper; Tezza, Alessandro; Haasdijk, Evert; Eiben, A. E.

Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018. ed. / Suresh Sundaram. Institute of Electrical and Electronics Engineers Inc., 2019. p. 851-858 8628857.

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Benefits of Social Learning in Physical Robots

AU - Heinerman, Jacqueline

AU - Bussmann, Bart

AU - Groenendijk, Rick

AU - Krieken, Emile Van

AU - Slik, Jesper

AU - Tezza, Alessandro

AU - Haasdijk, Evert

AU - Eiben, A. E.

PY - 2019/1/28

Y1 - 2019/1/28

N2 - Robot-to-robot learning, a specific case of social learning in robotics, enables the ability to transfer robot controllers directly from one robot to another. Previous studies showed that the exchange of controller information can increase learning speed and performance. However, most of these studies have been performed in simulation, where robots are identical. Therefore, the results do not necessarily transfer to a real environment, where each robot is unique per definition due to the random differences in hardware. In this paper, we investigate the effect of exchanging controller information, on top of individual learning, in a group of Thymio II robots for two tasks: obstacle avoidance and foraging. The controllers of the robots are neural networks that evolve using a modified version of the state-of-the-art NEAT algorithm, called cNEAT, which allows the conversion of innovations numbers from other robots. This paper shows that robot-to-robot learning seems to at least parallelise the search, reducing wall clock time. Additionally, controllers are less complex, resulting in a smaller search space.

AB - Robot-to-robot learning, a specific case of social learning in robotics, enables the ability to transfer robot controllers directly from one robot to another. Previous studies showed that the exchange of controller information can increase learning speed and performance. However, most of these studies have been performed in simulation, where robots are identical. Therefore, the results do not necessarily transfer to a real environment, where each robot is unique per definition due to the random differences in hardware. In this paper, we investigate the effect of exchanging controller information, on top of individual learning, in a group of Thymio II robots for two tasks: obstacle avoidance and foraging. The controllers of the robots are neural networks that evolve using a modified version of the state-of-the-art NEAT algorithm, called cNEAT, which allows the conversion of innovations numbers from other robots. This paper shows that robot-to-robot learning seems to at least parallelise the search, reducing wall clock time. Additionally, controllers are less complex, resulting in a smaller search space.

KW - Evolutionary Algorithms

KW - Evolutionary Robotics

KW - Neural Networks

KW - Robot-to-Robot Learning

UR - http://www.scopus.com/inward/record.url?scp=85062778438&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062778438&partnerID=8YFLogxK

U2 - 10.1109/SSCI.2018.8628857

DO - 10.1109/SSCI.2018.8628857

M3 - Conference contribution

SP - 851

EP - 858

BT - Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018

A2 - Sundaram, Suresh

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Heinerman J, Bussmann B, Groenendijk R, Krieken EV, Slik J, Tezza A et al. Benefits of Social Learning in Physical Robots. In Sundaram S, editor, Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018. Institute of Electrical and Electronics Engineers Inc. 2019. p. 851-858. 8628857 https://doi.org/10.1109/SSCI.2018.8628857