Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm

Jekanthan Thangavelautham, Nader Abu El Samid, Paul Grouchy, Ernest Earon, Terence Fu, Nagina Nagrani, Gabriele M T D'Eleuterio

Research output: Chapter in Book/Report/Conference proceedingConference contribution

10 Citations (Scopus)

Abstract

Autonomous robotic excavation has often been limited to a single robotic platform using a specified excavation vehicle. This paper presents a novel method for developing scalable controllers for use in multirobot scenarios and that do not require human defined operations scripts nor extensive modeling of the kinematics and dynamics of the excavation vehicles. Furthermore, the control system does not require specifying an excavation vehicle type such as a bulldozer, frontloader or bucket-wheel and it can evolve to select for an appropriate choice of excavation vehicles to successfully complete a task. The "Artificial Neural Tissue" (ANT) architecture is used as a control system for autonomous multirobot excavation and clearing tasks. This control architecture combines a variable-topology neural-network structure with a coarse-coding strategy that permits specialized areas to develop in the tissue. Training is done in a low-fidelity grid world simulation environment and where a single global fitness function and a set of allowable basis behaviors need be specified. This approach is found to provide improved training performance over fixed-topology neural networks and can be easily ported onto different robot platforms. Aspects of the controller functionality have been tested using high fidelity dynamics simulation and in hardware. An evolutionary training process discovers novel decentralized methods of cooperation employing aggregation behaviors (via synchronized movements). These aggregation behaviors are found to improve controller scalability (with increasing robot density) and better handle robot interference (antagonism) that reduces the overall efficiency of the group.

Original languageEnglish (US)
Title of host publicationProceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA
Pages258-265
Number of pages8
DOIs
StatePublished - 2009
Externally publishedYes
Event2009 IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA 2009 - Daejeon, Korea, Republic of
Duration: Dec 15 2009Dec 18 2009

Other

Other2009 IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA 2009
CountryKorea, Republic of
CityDaejeon
Period12/15/0912/18/09

Fingerprint

Multi-robot
Excavation
Robot
Paradigm
Tissue
Controller
Fidelity
Controllers
Robotics
Aggregation
Control System
Neural Networks
Topology
Antagonism
Robots
Simulation Environment
Fitness Function
Network Structure
Dynamic Simulation
Wheel

ASJC Scopus subject areas

  • Computational Mathematics

Cite this

Thangavelautham, J., El Samid, N. A., Grouchy, P., Earon, E., Fu, T., Nagrani, N., & D'Eleuterio, G. M. T. (2009). Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm. In Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA (pp. 258-265). [5423196] https://doi.org/10.1109/CIRA.2009.5423196

Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm. / Thangavelautham, Jekanthan; El Samid, Nader Abu; Grouchy, Paul; Earon, Ernest; Fu, Terence; Nagrani, Nagina; D'Eleuterio, Gabriele M T.

Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA. 2009. p. 258-265 5423196.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Thangavelautham, J, El Samid, NA, Grouchy, P, Earon, E, Fu, T, Nagrani, N & D'Eleuterio, GMT 2009, Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm. in Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA., 5423196, pp. 258-265, 2009 IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA 2009, Daejeon, Korea, Republic of, 12/15/09. https://doi.org/10.1109/CIRA.2009.5423196
Thangavelautham J, El Samid NA, Grouchy P, Earon E, Fu T, Nagrani N et al. Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm. In Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA. 2009. p. 258-265. 5423196 https://doi.org/10.1109/CIRA.2009.5423196
Thangavelautham, Jekanthan ; El Samid, Nader Abu ; Grouchy, Paul ; Earon, Ernest ; Fu, Terence ; Nagrani, Nagina ; D'Eleuterio, Gabriele M T. / Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm. Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA. 2009. pp. 258-265
@inproceedings{e05881eb026a4a55be933d114e11f5ac,
title = "Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm",
abstract = "Autonomous robotic excavation has often been limited to a single robotic platform using a specified excavation vehicle. This paper presents a novel method for developing scalable controllers for use in multirobot scenarios and that do not require human defined operations scripts nor extensive modeling of the kinematics and dynamics of the excavation vehicles. Furthermore, the control system does not require specifying an excavation vehicle type such as a bulldozer, frontloader or bucket-wheel and it can evolve to select for an appropriate choice of excavation vehicles to successfully complete a task. The {"}Artificial Neural Tissue{"} (ANT) architecture is used as a control system for autonomous multirobot excavation and clearing tasks. This control architecture combines a variable-topology neural-network structure with a coarse-coding strategy that permits specialized areas to develop in the tissue. Training is done in a low-fidelity grid world simulation environment and where a single global fitness function and a set of allowable basis behaviors need be specified. This approach is found to provide improved training performance over fixed-topology neural networks and can be easily ported onto different robot platforms. Aspects of the controller functionality have been tested using high fidelity dynamics simulation and in hardware. An evolutionary training process discovers novel decentralized methods of cooperation employing aggregation behaviors (via synchronized movements). These aggregation behaviors are found to improve controller scalability (with increasing robot density) and better handle robot interference (antagonism) that reduces the overall efficiency of the group.",
author = "Jekanthan Thangavelautham and {El Samid}, {Nader Abu} and Paul Grouchy and Ernest Earon and Terence Fu and Nagina Nagrani and D'Eleuterio, {Gabriele M T}",
year = "2009",
doi = "10.1109/CIRA.2009.5423196",
language = "English (US)",
pages = "258--265",
booktitle = "Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA",

}

TY - GEN

T1 - Evolving multirobot excavation controllers and choice of platforms using an artificial neural tissue paradigm

AU - Thangavelautham, Jekanthan

AU - El Samid, Nader Abu

AU - Grouchy, Paul

AU - Earon, Ernest

AU - Fu, Terence

AU - Nagrani, Nagina

AU - D'Eleuterio, Gabriele M T

PY - 2009

Y1 - 2009

N2 - Autonomous robotic excavation has often been limited to a single robotic platform using a specified excavation vehicle. This paper presents a novel method for developing scalable controllers for use in multirobot scenarios and that do not require human defined operations scripts nor extensive modeling of the kinematics and dynamics of the excavation vehicles. Furthermore, the control system does not require specifying an excavation vehicle type such as a bulldozer, frontloader or bucket-wheel and it can evolve to select for an appropriate choice of excavation vehicles to successfully complete a task. The "Artificial Neural Tissue" (ANT) architecture is used as a control system for autonomous multirobot excavation and clearing tasks. This control architecture combines a variable-topology neural-network structure with a coarse-coding strategy that permits specialized areas to develop in the tissue. Training is done in a low-fidelity grid world simulation environment and where a single global fitness function and a set of allowable basis behaviors need be specified. This approach is found to provide improved training performance over fixed-topology neural networks and can be easily ported onto different robot platforms. Aspects of the controller functionality have been tested using high fidelity dynamics simulation and in hardware. An evolutionary training process discovers novel decentralized methods of cooperation employing aggregation behaviors (via synchronized movements). These aggregation behaviors are found to improve controller scalability (with increasing robot density) and better handle robot interference (antagonism) that reduces the overall efficiency of the group.

AB - Autonomous robotic excavation has often been limited to a single robotic platform using a specified excavation vehicle. This paper presents a novel method for developing scalable controllers for use in multirobot scenarios and that do not require human defined operations scripts nor extensive modeling of the kinematics and dynamics of the excavation vehicles. Furthermore, the control system does not require specifying an excavation vehicle type such as a bulldozer, frontloader or bucket-wheel and it can evolve to select for an appropriate choice of excavation vehicles to successfully complete a task. The "Artificial Neural Tissue" (ANT) architecture is used as a control system for autonomous multirobot excavation and clearing tasks. This control architecture combines a variable-topology neural-network structure with a coarse-coding strategy that permits specialized areas to develop in the tissue. Training is done in a low-fidelity grid world simulation environment and where a single global fitness function and a set of allowable basis behaviors need be specified. This approach is found to provide improved training performance over fixed-topology neural networks and can be easily ported onto different robot platforms. Aspects of the controller functionality have been tested using high fidelity dynamics simulation and in hardware. An evolutionary training process discovers novel decentralized methods of cooperation employing aggregation behaviors (via synchronized movements). These aggregation behaviors are found to improve controller scalability (with increasing robot density) and better handle robot interference (antagonism) that reduces the overall efficiency of the group.

UR - http://www.scopus.com/inward/record.url?scp=77951188743&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77951188743&partnerID=8YFLogxK

U2 - 10.1109/CIRA.2009.5423196

DO - 10.1109/CIRA.2009.5423196

M3 - Conference contribution

SP - 258

EP - 265

BT - Proceedings of IEEE International Symposium on Computational Intelligence in Robotics and Automation, CIRA

ER -