Adaptive Strategy for Online Gait Learning Evaluated on the Polymorphic Robotic LocoKit

David Johan Christensen; Jørgen Christian Larsen; Kasper Stoy

doi:10.1109/EAIS.2012.6232806

Adaptive Strategy for Online Gait Learning Evaluated on the Polymorphic Robotic LocoKit

David Johan Christensen, Jørgen Christian Larsen, Kasper Stoy

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

500 Downloads (Pure)

Abstract

This paper presents experiments with a morphologyindependent, life-long strategy for online learning of locomotion gaits, performed on a quadruped robot constructed from the LocoKit modular robot. The learning strategy applies a stochastic optimization algorithm to optimize eight open parameters of a
central pattern generator based gait implementation. We observe that the strategy converges in roughly ten minutes to gaits of similar or higher velocity than a manually designed gait and that the strategy readapts in the event of failed actuators. In future work we plan to study co-learning of morphological and
control parameters directly on the physical robot.

Original language	English
Title of host publication	Proceedings of the IEEE Conference on Evolving and Adaptive Intelligent Systems
Number of pages	6
Publisher	IEEE
Publication date	2012
ISBN (Print)	9781467317283
DOIs	https://doi.org/10.1109/EAIS.2012.6232806
Publication status	Published - 2012
Event	2012 IEEE Conference on Evolving and Adaptive Intelligent Systems - Madrid, Spain Duration: 17 May 2012 → 18 May 2012

Conference

Conference	2012 IEEE Conference on Evolving and Adaptive Intelligent Systems
Country	Spain
City	Madrid
Period	17/05/2012 → 18/05/2012

Keywords

Actuators
Biology
Legged locomotion
Manuals.

Access to Document

10.1109/EAIS.2012.6232806

ieeeeais2012_submission_7.pdf

Fingerprint Dive into the research topics of 'Adaptive Strategy for Online Gait Learning Evaluated on the Polymorphic Robotic LocoKit'. Together they form a unique fingerprint.

Cite this

@inproceedings{a8c823dff676431fa525086957fd735f,

title = "Adaptive Strategy for Online Gait Learning Evaluated on the Polymorphic Robotic LocoKit",

abstract = "This paper presents experiments with a morphologyindependent, life-long strategy for online learning of locomotion gaits, performed on a quadruped robot constructed from the LocoKit modular robot. The learning strategy applies a stochastic optimization algorithm to optimize eight open parameters of acentral pattern generator based gait implementation. We observe that the strategy converges in roughly ten minutes to gaits of similar or higher velocity than a manually designed gait and that the strategy readapts in the event of failed actuators. In future work we plan to study co-learning of morphological andcontrol parameters directly on the physical robot.",

keywords = "Actuators, Biology, Legged locomotion, Manuals.",

author = "Christensen, {David Johan} and Larsen, {J{\o}rgen Christian} and Kasper Stoy",

year = "2012",

doi = "10.1109/EAIS.2012.6232806",

language = "English",

isbn = "9781467317283",

booktitle = "Proceedings of the IEEE Conference on Evolving and Adaptive Intelligent Systems",

publisher = "IEEE",

address = "United States",

note = "2012 IEEE Conference on Evolving and Adaptive Intelligent Systems, EAIS 2012 ; Conference date: 17-05-2012 Through 18-05-2012",

}

TY - GEN

T1 - Adaptive Strategy for Online Gait Learning Evaluated on the Polymorphic Robotic LocoKit

AU - Christensen, David Johan

AU - Larsen, Jørgen Christian

AU - Stoy, Kasper

PY - 2012

Y1 - 2012

N2 - This paper presents experiments with a morphologyindependent, life-long strategy for online learning of locomotion gaits, performed on a quadruped robot constructed from the LocoKit modular robot. The learning strategy applies a stochastic optimization algorithm to optimize eight open parameters of acentral pattern generator based gait implementation. We observe that the strategy converges in roughly ten minutes to gaits of similar or higher velocity than a manually designed gait and that the strategy readapts in the event of failed actuators. In future work we plan to study co-learning of morphological andcontrol parameters directly on the physical robot.

AB - This paper presents experiments with a morphologyindependent, life-long strategy for online learning of locomotion gaits, performed on a quadruped robot constructed from the LocoKit modular robot. The learning strategy applies a stochastic optimization algorithm to optimize eight open parameters of acentral pattern generator based gait implementation. We observe that the strategy converges in roughly ten minutes to gaits of similar or higher velocity than a manually designed gait and that the strategy readapts in the event of failed actuators. In future work we plan to study co-learning of morphological andcontrol parameters directly on the physical robot.

KW - Actuators

KW - Biology

KW - Legged locomotion

KW - Manuals.

U2 - 10.1109/EAIS.2012.6232806

DO - 10.1109/EAIS.2012.6232806

M3 - Article in proceedings

SN - 9781467317283

BT - Proceedings of the IEEE Conference on Evolving and Adaptive Intelligent Systems

PB - IEEE

T2 - 2012 IEEE Conference on Evolving and Adaptive Intelligent Systems

Y2 - 17 May 2012 through 18 May 2012

ER -