Control of a simulated arm using a novel combination of a cerebellar learning mechanisms

Christopher Assad; Mitra J. Hartmann; Michael G. Paulin

doi:10.1016/S0925-2312(02)00351-X

Control of a simulated arm using a novel combination of a cerebellar learning mechanisms

Christopher Assad^*, Mitra J. Hartmann, Michael G. Paulin

^*Corresponding author for this work

Biomedical Engineering

Research output: Contribution to journal › Article › peer-review

1 Scopus citations

Abstract

We present a model of cerebellar cortex that combines two types of learning: feedforward predictive association based on local Hebbian-type learning between granule cell ascending branch and parallel fiber inputs, and reinforcement learning with feedback error correction based on climbing fiber activity. The model is motivated by recent physiological and anatomical evidence and has more computational capacity than previous functional models of cerebellum. To demonstrate the model's utility, we simulated the control of a simple virtual arm. The model successfully learned to control the timing of release for the arm during a target-throwing task.

Original language	English (US)
Pages (from-to)	275-283
Number of pages	9
Journal	Neurocomputing
Volume	44-46
DOIs	https://doi.org/10.1016/S0925-2312(02)00351-X
State	Published - Jun 1 2002

Funding

The research described in this paper was performed at the Center for Integrated Space Microsystems, Jet Propulsion Laboratory, California Institute of Technology, and was sponsored by the National Aeronautics and Space Administration.

Keywords

Cerebellar learning
Cerebellum
Dynamic state estimation
Sensorimotor control

ASJC Scopus subject areas

Computer Science Applications
Cognitive Neuroscience
Artificial Intelligence

Access to Document

10.1016/S0925-2312(02)00351-X

Cite this

@article{9a2703109a814f86a76849359f6e7159,

title = "Control of a simulated arm using a novel combination of a cerebellar learning mechanisms",

abstract = "We present a model of cerebellar cortex that combines two types of learning: feedforward predictive association based on local Hebbian-type learning between granule cell ascending branch and parallel fiber inputs, and reinforcement learning with feedback error correction based on climbing fiber activity. The model is motivated by recent physiological and anatomical evidence and has more computational capacity than previous functional models of cerebellum. To demonstrate the model's utility, we simulated the control of a simple virtual arm. The model successfully learned to control the timing of release for the arm during a target-throwing task.",

keywords = "Cerebellar learning, Cerebellum, Dynamic state estimation, Sensorimotor control",

author = "Christopher Assad and Hartmann, {Mitra J.} and Paulin, {Michael G.}",

note = "Funding Information: The research described in this paper was performed at the Center for Integrated Space Microsystems, Jet Propulsion Laboratory, California Institute of Technology, and was sponsored by the National Aeronautics and Space Administration.",

year = "2002",

month = jun,

day = "1",

doi = "10.1016/S0925-2312(02)00351-X",

language = "English (US)",

volume = "44-46",

pages = "275--283",

journal = "Neurocomputing",

issn = "0925-2312",

publisher = "Elsevier B.V.",

}

TY - JOUR

T1 - Control of a simulated arm using a novel combination of a cerebellar learning mechanisms

AU - Assad, Christopher

AU - Hartmann, Mitra J.

AU - Paulin, Michael G.

N1 - Funding Information: The research described in this paper was performed at the Center for Integrated Space Microsystems, Jet Propulsion Laboratory, California Institute of Technology, and was sponsored by the National Aeronautics and Space Administration.

PY - 2002/6/1

Y1 - 2002/6/1

N2 - We present a model of cerebellar cortex that combines two types of learning: feedforward predictive association based on local Hebbian-type learning between granule cell ascending branch and parallel fiber inputs, and reinforcement learning with feedback error correction based on climbing fiber activity. The model is motivated by recent physiological and anatomical evidence and has more computational capacity than previous functional models of cerebellum. To demonstrate the model's utility, we simulated the control of a simple virtual arm. The model successfully learned to control the timing of release for the arm during a target-throwing task.

AB - We present a model of cerebellar cortex that combines two types of learning: feedforward predictive association based on local Hebbian-type learning between granule cell ascending branch and parallel fiber inputs, and reinforcement learning with feedback error correction based on climbing fiber activity. The model is motivated by recent physiological and anatomical evidence and has more computational capacity than previous functional models of cerebellum. To demonstrate the model's utility, we simulated the control of a simple virtual arm. The model successfully learned to control the timing of release for the arm during a target-throwing task.

KW - Cerebellar learning

KW - Cerebellum

KW - Dynamic state estimation

KW - Sensorimotor control

UR - http://www.scopus.com/inward/record.url?scp=0036067479&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0036067479&partnerID=8YFLogxK

U2 - 10.1016/S0925-2312(02)00351-X

DO - 10.1016/S0925-2312(02)00351-X

M3 - Article

AN - SCOPUS:0036067479

SN - 0925-2312

VL - 44-46

SP - 275

EP - 283

JO - Neurocomputing

JF - Neurocomputing

ER -

Control of a simulated arm using a novel combination of a cerebellar learning mechanisms

Abstract

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this