Evolutionary vs imitation learning for neuromorphic control at the edge

Catherine Schuman; Robert Patton; Shruti Kulkarni; Maryam Parsa; Christopher Stahl; N. Quentin Haas; J. Parker Mitchell; Shay Snyder; Amelie Nagle; Alexandra Shanafield; Thomas Potok

doi:10.1088/2634-4386/ac45e7

Evolutionary vs imitation learning for neuromorphic control at the edge

Catherine Schuman, Robert Patton, Shruti Kulkarni, Maryam Parsa, Christopher Stahl, N. Quentin Haas, J. Parker Mitchell, Shay Snyder, Amelie Nagle, Alexandra Shanafield, Thomas Potok

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

14 Scopus citations

Abstract

Neuromorphic computing offers the opportunity to implement extremely low power artificial intelligence at the edge. Control applications, such as autonomous vehicles and robotics, are also of great interest for neuromorphic systems at the edge. It is not clear, however, what the best neuromorphic training approaches are for control applications at the edge. In this work, we implement and compare the performance of evolutionary optimization and imitation learning approaches on an autonomous race car control task using an edge neuromorphic implementation. We show that the evolutionary approaches tend to achieve better performing smaller network sizes that are well-suited to edge deployment, but they also take significantly longer to train. We also describe a workflow to allow for future algorithmic comparisons for neuromorphic hardware on control applications at the edge.

Original language	English
Article number	014002
Journal	Neuromorphic Computing and Engineering
Volume	2
Issue number	1
DOIs	https://doi.org/10.1088/2634-4386/ac45e7
State	Published - Mar 1 2022

Funding

Notice: this manuscript has been authored in part by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan ( http://energy.gov/downloads/doe-public-access-plan ). This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research, under Contract number DE-AC05-00OR22725.

Keywords

autonomous driving
evolutionary optimization
imitation learning
spiking neural networks

Access to Document

10.1088/2634-4386/ac45e7

Cite this

@article{443dd8d004f74201b6255149bd8c1e79,

title = "Evolutionary vs imitation learning for neuromorphic control at the edge",

abstract = "Neuromorphic computing offers the opportunity to implement extremely low power artificial intelligence at the edge. Control applications, such as autonomous vehicles and robotics, are also of great interest for neuromorphic systems at the edge. It is not clear, however, what the best neuromorphic training approaches are for control applications at the edge. In this work, we implement and compare the performance of evolutionary optimization and imitation learning approaches on an autonomous race car control task using an edge neuromorphic implementation. We show that the evolutionary approaches tend to achieve better performing smaller network sizes that are well-suited to edge deployment, but they also take significantly longer to train. We also describe a workflow to allow for future algorithmic comparisons for neuromorphic hardware on control applications at the edge.",

keywords = "autonomous driving, evolutionary optimization, imitation learning, spiking neural networks",

author = "Catherine Schuman and Robert Patton and Shruti Kulkarni and Maryam Parsa and Christopher Stahl and Haas, {N. Quentin} and Mitchell, {J. Parker} and Shay Snyder and Amelie Nagle and Alexandra Shanafield and Thomas Potok",

note = "Publisher Copyright: {\textcopyright} 2022 The Author(s). Published by IOP Publishing Ltd.",

year = "2022",

month = mar,

day = "1",

doi = "10.1088/2634-4386/ac45e7",

language = "English",

volume = "2",

journal = "Neuromorphic Computing and Engineering",

issn = "2634-4386",

publisher = "IOP Publishing",

number = "1",

}

TY - JOUR

T1 - Evolutionary vs imitation learning for neuromorphic control at the edge

AU - Schuman, Catherine

AU - Patton, Robert

AU - Kulkarni, Shruti

AU - Parsa, Maryam

AU - Stahl, Christopher

AU - Haas, N. Quentin

AU - Mitchell, J. Parker

AU - Snyder, Shay

AU - Nagle, Amelie

AU - Shanafield, Alexandra

AU - Potok, Thomas

PY - 2022/3/1

Y1 - 2022/3/1

N2 - Neuromorphic computing offers the opportunity to implement extremely low power artificial intelligence at the edge. Control applications, such as autonomous vehicles and robotics, are also of great interest for neuromorphic systems at the edge. It is not clear, however, what the best neuromorphic training approaches are for control applications at the edge. In this work, we implement and compare the performance of evolutionary optimization and imitation learning approaches on an autonomous race car control task using an edge neuromorphic implementation. We show that the evolutionary approaches tend to achieve better performing smaller network sizes that are well-suited to edge deployment, but they also take significantly longer to train. We also describe a workflow to allow for future algorithmic comparisons for neuromorphic hardware on control applications at the edge.

AB - Neuromorphic computing offers the opportunity to implement extremely low power artificial intelligence at the edge. Control applications, such as autonomous vehicles and robotics, are also of great interest for neuromorphic systems at the edge. It is not clear, however, what the best neuromorphic training approaches are for control applications at the edge. In this work, we implement and compare the performance of evolutionary optimization and imitation learning approaches on an autonomous race car control task using an edge neuromorphic implementation. We show that the evolutionary approaches tend to achieve better performing smaller network sizes that are well-suited to edge deployment, but they also take significantly longer to train. We also describe a workflow to allow for future algorithmic comparisons for neuromorphic hardware on control applications at the edge.

KW - autonomous driving

KW - evolutionary optimization

KW - imitation learning

KW - spiking neural networks

UR - http://www.scopus.com/inward/record.url?scp=85130050301&partnerID=8YFLogxK

U2 - 10.1088/2634-4386/ac45e7

DO - 10.1088/2634-4386/ac45e7

M3 - Article

AN - SCOPUS:85130050301

SN - 2634-4386

VL - 2

JO - Neuromorphic Computing and Engineering

JF - Neuromorphic Computing and Engineering

IS - 1

M1 - 014002

ER -

Evolutionary vs imitation learning for neuromorphic control at the edge

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this