Neural network-based learning from demonstration of an autonomous ground robot

Yiwei Fu, Devesh K. Jha, Zeyu Zhang, Zhenyuan Yuan, Asok Ray

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

This paper presents and experimentally validates a concept of end-to-end imitation learning for autonomous systems by using a composite architecture of convolutional neural network (ConvNet) and Long Short Term Memory (LSTM) neural network. In particular, a spatio-temporal deep neural network is developed, which learns to imitate the policy used by a human supervisor to drive a car-like robot in a maze environment. The spatial and temporal components of the imitation model are learned by using deep convolutional network and recurrent neural network architectures, respectively. The imitation model learns the policy of a human supervisor as a function of laser light detection and ranging (LIDAR) data, which is then used in real time to drive a robot in an autonomous fashion in a laboratory setting. The performance of the proposed model for imitation learning is compared with that of several other state-of-the-art methods, reported in the machine learning literature, for spatial and temporal modeling. The learned policy is implemented on a robot using a Nvidia Jetson TX2 board which, in turn, is validated on test tracks. The proposed spatio-temporal model outperforms several other off-the-shelf machine learning techniques to learn the policy.

Original languageEnglish (US)
Article number24
JournalMachines
Volume7
Issue number2
DOIs
StatePublished - Jun 1 2019

Fingerprint

Imitation
Demonstrations
Robot
Robots
Neural Networks
Neural networks
Supervisory personnel
Learning systems
Machine Learning
Spatio-temporal Model
Memory Term
Recurrent neural networks
Recurrent Neural Networks
Network Architecture
Autonomous Systems
Network architecture
Railroad cars
Composite
Model
Laser

All Science Journal Classification (ASJC) codes

  • Computer Science (miscellaneous)
  • Control and Systems Engineering
  • Mechanical Engineering
  • Control and Optimization
  • Industrial and Manufacturing Engineering
  • Electrical and Electronic Engineering

Cite this

Fu, Yiwei ; Jha, Devesh K. ; Zhang, Zeyu ; Yuan, Zhenyuan ; Ray, Asok. / Neural network-based learning from demonstration of an autonomous ground robot. In: Machines. 2019 ; Vol. 7, No. 2.
@article{6dc0f9bc660349bea5ae7f18b4ebb6cb,
title = "Neural network-based learning from demonstration of an autonomous ground robot",
abstract = "This paper presents and experimentally validates a concept of end-to-end imitation learning for autonomous systems by using a composite architecture of convolutional neural network (ConvNet) and Long Short Term Memory (LSTM) neural network. In particular, a spatio-temporal deep neural network is developed, which learns to imitate the policy used by a human supervisor to drive a car-like robot in a maze environment. The spatial and temporal components of the imitation model are learned by using deep convolutional network and recurrent neural network architectures, respectively. The imitation model learns the policy of a human supervisor as a function of laser light detection and ranging (LIDAR) data, which is then used in real time to drive a robot in an autonomous fashion in a laboratory setting. The performance of the proposed model for imitation learning is compared with that of several other state-of-the-art methods, reported in the machine learning literature, for spatial and temporal modeling. The learned policy is implemented on a robot using a Nvidia Jetson TX2 board which, in turn, is validated on test tracks. The proposed spatio-temporal model outperforms several other off-the-shelf machine learning techniques to learn the policy.",
author = "Yiwei Fu and Jha, {Devesh K.} and Zeyu Zhang and Zhenyuan Yuan and Asok Ray",
year = "2019",
month = "6",
day = "1",
doi = "10.3390/machines7020024",
language = "English (US)",
volume = "7",
journal = "Machines",
issn = "2075-1702",
publisher = "MDPI AG",
number = "2",

}

Neural network-based learning from demonstration of an autonomous ground robot. / Fu, Yiwei; Jha, Devesh K.; Zhang, Zeyu; Yuan, Zhenyuan; Ray, Asok.

In: Machines, Vol. 7, No. 2, 24, 01.06.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Neural network-based learning from demonstration of an autonomous ground robot

AU - Fu, Yiwei

AU - Jha, Devesh K.

AU - Zhang, Zeyu

AU - Yuan, Zhenyuan

AU - Ray, Asok

PY - 2019/6/1

Y1 - 2019/6/1

N2 - This paper presents and experimentally validates a concept of end-to-end imitation learning for autonomous systems by using a composite architecture of convolutional neural network (ConvNet) and Long Short Term Memory (LSTM) neural network. In particular, a spatio-temporal deep neural network is developed, which learns to imitate the policy used by a human supervisor to drive a car-like robot in a maze environment. The spatial and temporal components of the imitation model are learned by using deep convolutional network and recurrent neural network architectures, respectively. The imitation model learns the policy of a human supervisor as a function of laser light detection and ranging (LIDAR) data, which is then used in real time to drive a robot in an autonomous fashion in a laboratory setting. The performance of the proposed model for imitation learning is compared with that of several other state-of-the-art methods, reported in the machine learning literature, for spatial and temporal modeling. The learned policy is implemented on a robot using a Nvidia Jetson TX2 board which, in turn, is validated on test tracks. The proposed spatio-temporal model outperforms several other off-the-shelf machine learning techniques to learn the policy.

AB - This paper presents and experimentally validates a concept of end-to-end imitation learning for autonomous systems by using a composite architecture of convolutional neural network (ConvNet) and Long Short Term Memory (LSTM) neural network. In particular, a spatio-temporal deep neural network is developed, which learns to imitate the policy used by a human supervisor to drive a car-like robot in a maze environment. The spatial and temporal components of the imitation model are learned by using deep convolutional network and recurrent neural network architectures, respectively. The imitation model learns the policy of a human supervisor as a function of laser light detection and ranging (LIDAR) data, which is then used in real time to drive a robot in an autonomous fashion in a laboratory setting. The performance of the proposed model for imitation learning is compared with that of several other state-of-the-art methods, reported in the machine learning literature, for spatial and temporal modeling. The learned policy is implemented on a robot using a Nvidia Jetson TX2 board which, in turn, is validated on test tracks. The proposed spatio-temporal model outperforms several other off-the-shelf machine learning techniques to learn the policy.

UR - http://www.scopus.com/inward/record.url?scp=85069052252&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85069052252&partnerID=8YFLogxK

U2 - 10.3390/machines7020024

DO - 10.3390/machines7020024

M3 - Article

AN - SCOPUS:85069052252

VL - 7

JO - Machines

JF - Machines

SN - 2075-1702

IS - 2

M1 - 24

ER -