DocumentCode :
3307154
Title :
Customized learning algorithms for episodic tasks with acyclic state spaces
Author :
Bountourelis, Theologos ; Reveliotis, Spyros
Author_Institution :
Sch. of Ind. & Syst. Eng., Georgia Inst. of Technol., Atlanta, GA, USA
fYear :
2009
fDate :
22-25 Aug. 2009
Firstpage :
627
Lastpage :
634
Abstract :
The work presented in this paper provides a practical, customized learning algorithm for reinforcement learning tasks that evolve episodically over acyclic state spaces. The presented results are motivated by the optimal disassembly planning (ODP) problem described in, and they complement and enhance some earlier developments on this problem that were presented in. In particular, the proposed algorithm is shown to be a substantial improvement of the original algorithm developed in, in terms of, both, the involved computational effort and the attained performance, where the latter is measured by the accumulated reward. The new algorithm also leads to a robust performance gain over the typical Q-learning implementations for the considered problem context.
Keywords :
learning (artificial intelligence); Q-learning implementation; acyclic state space; customized learning algorithm; episodic task; optimal disassembly planning; reinforcement learning; robust performance gain; Aerospace industry; Algorithm design and analysis; Automation; Convergence; Data mining; Learning; Q factor; Space technology; State-space methods; Systems engineering and theory;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Automation Science and Engineering, 2009. CASE 2009. IEEE International Conference on
Conference_Location :
Bangalore
Print_ISBN :
978-1-4244-4578-3
Electronic_ISBN :
978-1-4244-4579-0
Type :
conf
DOI :
10.1109/COASE.2009.5234189
Filename :
5234189
Link To Document :
بازگشت