Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework

U Kartoun; H Stem; Y Edan; C Feied; J Handler; M Smith; M Gillam

Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework

U Kartoun, H Stem, Y Edan, C Feied, J Handler, M Smith, M Gillam

Department of Industrial Engineering and Management

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

Abstract

This paper presents the design and implementation of a new reinforcement learning (RL) based algorithm. The proposed algorithm, CQ(λ) (collaborative Q(λ) ) allows several learning agents to acquire knowledge from each other. Acquiring knowledge learnt by an agent via collaboration with another agent enables acceleration of the entire learning system; therefore, learning can be utilized more efficiently. By developing collaborative learning algorithms, a learning task solution can be achieved significantly faster if performed by a single agent only, namely the number of learning episodes to solve a task is reduced. The proposed algorithm proved to accelerate learning in navigation robotic problem. The CQ(λ) algorithm was applied to autonomous mobile
robot navigation where several robot agents serve as learning processes. Robots learned to navigate an 11 x 11 world contains obstacles and boundaries choosing the optimum path to reach a target. Simulated experiments based on 50 learning episodes showed an average improvement of 17.02% while measuring the number of learning steps required reaching definite optimality and an average improvement of 32.98% for convergence to near
optimality by using two robots compared with the Q(λ) algorithm [1, 2].

Original language	English
Title of host publication	Proceedings of the sixth IASTED International Conference on Robotics and Applications
Subtitle of host publication	October 31-November 2, 2005, Cambridge, MA, USA
Editors	L. A. Gerhardt
Publisher	ACTA Press
Pages	13-19
ISBN (Print)	0889865213, 9780889865211
State	Published - 2005
Event	IASTED International Conference on Robotics and Applications - Cambridge, United States Duration: 31 Oct 2005 → 2 Nov 2005

Conference

Conference	IASTED International Conference on Robotics and Applications
Country/Territory	United States
City	Cambridge
Period	31/10/05 → 2/11/05

Keywords

robot simulation
reinforcement learning
and navigation

Access to Document

https://www.actapress.com/Abstract.aspx?paperId=22964

Cite this

Kartoun, U., Stem, H., Edan, Y., Feied, C., Handler, J., Smith, M., & Gillam, M. (2005). Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework. In L. A. Gerhardt (Ed.), Proceedings of the sixth IASTED International Conference on Robotics and Applications: October 31-November 2, 2005, Cambridge, MA, USA (pp. 13-19). ACTA Press. https://www.actapress.com/Abstract.aspx?paperId=22964

@inproceedings{356f7fb78d8c4e1fbb38cb5b03044533,

title = "Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework",

abstract = "This paper presents the design and implementation of a new reinforcement learning (RL) based algorithm. The proposed algorithm, CQ(λ) (collaborative Q(λ) ) allows several learning agents to acquire knowledge from each other. Acquiring knowledge learnt by an agent via collaboration with another agent enables acceleration of the entire learning system; therefore, learning can be utilized more efficiently. By developing collaborative learning algorithms, a learning task solution can be achieved significantly faster if performed by a single agent only, namely the number of learning episodes to solve a task is reduced. The proposed algorithm proved to accelerate learning in navigation robotic problem. The CQ(λ) algorithm was applied to autonomous mobilerobot navigation where several robot agents serve as learning processes. Robots learned to navigate an 11 x 11 world contains obstacles and boundaries choosing the optimum path to reach a target. Simulated experiments based on 50 learning episodes showed an average improvement of 17.02% while measuring the number of learning steps required reaching definite optimality and an average improvement of 32.98% for convergence to nearoptimality by using two robots compared with the Q(λ) algorithm [1, 2].",

keywords = "robot simulation, reinforcement learning, and navigation",

author = "U Kartoun and H Stem and Y Edan and C Feied and J Handler and M Smith and M Gillam",

year = "2005",

language = "English",

isbn = "0889865213",

pages = "13--19",

editor = "Gerhardt, {L. A. }",

booktitle = "Proceedings of the sixth IASTED International Conference on Robotics and Applications",

publisher = "ACTA Press",

address = "Canada",

note = "IASTED International Conference on Robotics and Applications ; Conference date: 31-10-2005 Through 02-11-2005",

}

Kartoun, U, Stem, H, Edan, Y, Feied, C, Handler, J, Smith, M & Gillam, M 2005, Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework. in LA Gerhardt (ed.), Proceedings of the sixth IASTED International Conference on Robotics and Applications: October 31-November 2, 2005, Cambridge, MA, USA. ACTA Press, pp. 13-19, IASTED International Conference on Robotics and Applications, Cambridge, Massachusetts, United States, 31/10/05. <https://www.actapress.com/Abstract.aspx?paperId=22964>

Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework. / Kartoun, U; Stem, H; Edan, Y et al.
Proceedings of the sixth IASTED International Conference on Robotics and Applications: October 31-November 2, 2005, Cambridge, MA, USA. ed. / L. A. Gerhardt. ACTA Press, 2005. p. 13-19.

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

TY - GEN

T1 - Collaborative Q(λ) Reinforcement Learning Algorithm

T2 - IASTED International Conference on Robotics and Applications

AU - Kartoun, U

AU - Stem, H

AU - Edan, Y

AU - Feied, C

AU - Handler, J

AU - Smith, M

AU - Gillam, M

PY - 2005

Y1 - 2005

N2 - This paper presents the design and implementation of a new reinforcement learning (RL) based algorithm. The proposed algorithm, CQ(λ) (collaborative Q(λ) ) allows several learning agents to acquire knowledge from each other. Acquiring knowledge learnt by an agent via collaboration with another agent enables acceleration of the entire learning system; therefore, learning can be utilized more efficiently. By developing collaborative learning algorithms, a learning task solution can be achieved significantly faster if performed by a single agent only, namely the number of learning episodes to solve a task is reduced. The proposed algorithm proved to accelerate learning in navigation robotic problem. The CQ(λ) algorithm was applied to autonomous mobilerobot navigation where several robot agents serve as learning processes. Robots learned to navigate an 11 x 11 world contains obstacles and boundaries choosing the optimum path to reach a target. Simulated experiments based on 50 learning episodes showed an average improvement of 17.02% while measuring the number of learning steps required reaching definite optimality and an average improvement of 32.98% for convergence to nearoptimality by using two robots compared with the Q(λ) algorithm [1, 2].

AB - This paper presents the design and implementation of a new reinforcement learning (RL) based algorithm. The proposed algorithm, CQ(λ) (collaborative Q(λ) ) allows several learning agents to acquire knowledge from each other. Acquiring knowledge learnt by an agent via collaboration with another agent enables acceleration of the entire learning system; therefore, learning can be utilized more efficiently. By developing collaborative learning algorithms, a learning task solution can be achieved significantly faster if performed by a single agent only, namely the number of learning episodes to solve a task is reduced. The proposed algorithm proved to accelerate learning in navigation robotic problem. The CQ(λ) algorithm was applied to autonomous mobilerobot navigation where several robot agents serve as learning processes. Robots learned to navigate an 11 x 11 world contains obstacles and boundaries choosing the optimum path to reach a target. Simulated experiments based on 50 learning episodes showed an average improvement of 17.02% while measuring the number of learning steps required reaching definite optimality and an average improvement of 32.98% for convergence to nearoptimality by using two robots compared with the Q(λ) algorithm [1, 2].

KW - robot simulation

KW - reinforcement learning

KW - and navigation

M3 - Conference contribution

SN - 0889865213

SN - 9780889865211

SP - 13

EP - 19

BT - Proceedings of the sixth IASTED International Conference on Robotics and Applications

A2 - Gerhardt, L. A.

PB - ACTA Press

Y2 - 31 October 2005 through 2 November 2005

ER -

Collaborative Q(λ) Reinforcement Learning Algorithm: A Promising Robot Learning Framework

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this