Learning in Growing Robots: Knowledge Transfer from Tadpole to Frog Robot

Yiheng Zhu; Jonathan Rossiter; Helmut Hauser

doi:10.1007/978-3-030-24741-6_42

Learning in Growing Robots: Knowledge Transfer from Tadpole to Frog Robot

Yiheng Zhu, Jonathan Rossiter, Helmut Hauser

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

3 Citations (Scopus)

182 Downloads (Pure)

Abstract

Inspired by natural growing processes, we investigate how morphological changes can potentially help to lead and facilitate the task of learning to control a robot. We use the model of a tadpole that grows in four discrete stages into a frog. The control task to learn is to locomote to food positions that occur at random positions. We employ reinforcement learning, which is able to find a tail-driven swimming strategy for the tadpole stage that transitions into a leg-driven strategy for the frog. Furthermore, by using knowledge transferred from one growing stage to the next one, we were able to show that growing can benefit from guiding the controller optimization through morphological changes. The results suggest that learning time can be reduced compared to the cases when learning each stage individually from scratch.

Original language	English
Title of host publication	Conference on Biomimetic and Biohybrid Systems
Subtitle of host publication	Living Machines 2019: Biomimetic and Biohybrid Systems
Publisher	Springer, Cham
Pages	378-382
Number of pages	5
ISBN (Print)	9783030247409
DOIs	https://doi.org/10.1007/978-3-030-24741-6_42
Publication status	Published - 6 Jul 2019
Event	2019 Living Machines - Kasugano International Forum, Nara, Japan Duration: 9 Jul 2019 → 12 Jul 2019 Conference number: 8 http://livingmachinesconference.eu/2019

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer Cham
Volume	11556

Conference

Conference	2019 Living Machines
Country/Territory	Japan
City	Nara
Period	9/07/19 → 12/07/19
Internet address	http://livingmachinesconference.eu/2019

Research Groups and Themes

Tactile Action Perception

Keywords

morphological computation
reinforcement learning
knowledge transfer
biomimetic robotics

Access to Document

10.1007/978-3-030-24741-6_42

Full-text PDF (accepted author manuscript)
This is the accepted author manuscript (AAM). The final published version (version of record) is available online via Springer Cham at https://doi.org/10.1007/978-3-030-24741-6_42 . Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 1.63 MBLicence: Other

Handle.net

Persistent link

Cite this

@inproceedings{5fa49acc179746d19075af3fd4677cd8,

title = "Learning in Growing Robots: Knowledge Transfer from Tadpole to Frog Robot",

abstract = "Inspired by natural growing processes, we investigate how morphological changes can potentially help to lead and facilitate the task of learning to control a robot. We use the model of a tadpole that grows in four discrete stages into a frog. The control task to learn is to locomote to food positions that occur at random positions. We employ reinforcement learning, which is able to find a tail-driven swimming strategy for the tadpole stage that transitions into a leg-driven strategy for the frog. Furthermore, by using knowledge transferred from one growing stage to the next one, we were able to show that growing can benefit from guiding the controller optimization through morphological changes. The results suggest that learning time can be reduced compared to the cases when learning each stage individually from scratch.",

keywords = "morphological computation, reinforcement learning, knowledge transfer, biomimetic robotics",

author = "Yiheng Zhu and Jonathan Rossiter and Helmut Hauser",

year = "2019",

month = jul,

day = "6",

doi = "10.1007/978-3-030-24741-6_42",

language = "English",

isbn = "9783030247409",

series = "Lecture Notes in Computer Science",

publisher = "Springer, Cham",

pages = "378--382",

booktitle = "Conference on Biomimetic and Biohybrid Systems",

address = "Switzerland",

note = "2019 Living Machines ; Conference date: 09-07-2019 Through 12-07-2019",

url = "http://livingmachinesconference.eu/2019",

}

Zhu, Y, Rossiter, J & Hauser, H 2019, Learning in Growing Robots: Knowledge Transfer from Tadpole to Frog Robot. in Conference on Biomimetic and Biohybrid Systems: Living Machines 2019: Biomimetic and Biohybrid Systems. Lecture Notes in Computer Science, vol. 11556, Springer, Cham, pp. 378-382, 2019 Living Machines, Nara, Japan, 9/07/19. https://doi.org/10.1007/978-3-030-24741-6_42

Learning in Growing Robots: Knowledge Transfer from Tadpole to Frog Robot. / Zhu, Yiheng; Rossiter, Jonathan ; Hauser, Helmut.
Conference on Biomimetic and Biohybrid Systems: Living Machines 2019: Biomimetic and Biohybrid Systems. Springer, Cham, 2019. p. 378-382 (Lecture Notes in Computer Science; Vol. 11556).

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Learning in Growing Robots

T2 - 2019 Living Machines

AU - Zhu, Yiheng

AU - Rossiter, Jonathan

AU - Hauser, Helmut

N1 - Conference code: 8

PY - 2019/7/6

Y1 - 2019/7/6

N2 - Inspired by natural growing processes, we investigate how morphological changes can potentially help to lead and facilitate the task of learning to control a robot. We use the model of a tadpole that grows in four discrete stages into a frog. The control task to learn is to locomote to food positions that occur at random positions. We employ reinforcement learning, which is able to find a tail-driven swimming strategy for the tadpole stage that transitions into a leg-driven strategy for the frog. Furthermore, by using knowledge transferred from one growing stage to the next one, we were able to show that growing can benefit from guiding the controller optimization through morphological changes. The results suggest that learning time can be reduced compared to the cases when learning each stage individually from scratch.

AB - Inspired by natural growing processes, we investigate how morphological changes can potentially help to lead and facilitate the task of learning to control a robot. We use the model of a tadpole that grows in four discrete stages into a frog. The control task to learn is to locomote to food positions that occur at random positions. We employ reinforcement learning, which is able to find a tail-driven swimming strategy for the tadpole stage that transitions into a leg-driven strategy for the frog. Furthermore, by using knowledge transferred from one growing stage to the next one, we were able to show that growing can benefit from guiding the controller optimization through morphological changes. The results suggest that learning time can be reduced compared to the cases when learning each stage individually from scratch.

KW - morphological computation

KW - reinforcement learning

KW - knowledge transfer

KW - biomimetic robotics

U2 - 10.1007/978-3-030-24741-6_42

DO - 10.1007/978-3-030-24741-6_42

M3 - Conference Contribution (Conference Proceeding)

SN - 9783030247409

T3 - Lecture Notes in Computer Science

SP - 378

EP - 382

BT - Conference on Biomimetic and Biohybrid Systems

PB - Springer, Cham

Y2 - 9 July 2019 through 12 July 2019

ER -

Learning in Growing Robots: Knowledge Transfer from Tadpole to Frog Robot

Abstract

Publication series

Conference

Research Groups and Themes

Keywords

Access to Document

Handle.net

Fingerprint

Cite this