A Policy-Improving System for Adaptability to Dynamic Environments Using Mixture Probability and Clustering DistributionReportar como inadecuado




A Policy-Improving System for Adaptability to Dynamic Environments Using Mixture Probability and Clustering Distribution - Descarga este documento en PDF. Documentación en PDF para descargar gratis. Disponible también para leer online.

Along with the increasing need for rescue robotsin disasters such as earthquakes and tsunami, there is an urgent need to developrobotics software for learning and adapting to any environment. A reinforcementlearning RL system that improves agents’ policies for dynamic environments byusing a mixture model of Bayesian networks has been proposed, and is effective inquickly adapting to a changing environment. However, the increase in computationalcomplexity requires the use of a high-performance computer for simulated experimentsand in the case of limited calculation resources, it becomes necessary to controlthe computational complexity. In this study, we used an RL profit-sharing methodfor the agent to learn its policy, and introduced a mixture probability into theRL system to recognize changes in the environment and appropriately improve theagent’s policy to adjust to a changing environment. We also introduced a clusteringdistribution that enables a smaller, suitable selection, while maintaining a varietyof mixture probability elements in order to reduce the computational complexityand simultaneously maintain the system’s performance. Using our proposed system,the agent successfully learned the policy and efficiently adjusted to the changingenvironment. Finally, control of the computational complexity was effective, andthe decline in effectiveness of the policy improvement was controlled by using ourproposed system.

KEYWORDS

Reinforcement Learning; Profit-Sharing Method; Mixture Probability; Clustering

Cite this paper

Phommasak, U. , Kitakoshi, D. , Mao, J. and Shioya, H. 2014 A Policy-Improving System for Adaptability to Dynamic Environments Using Mixture Probability and Clustering Distribution. Journal of Computer and Communications, 2, 210-219. doi: 10.4236-jcc.2014.24028.





Autor: Uthai Phommasak, Daisuke Kitakoshi, Jun Mao, Hiroyuki Shioya

Fuente: http://www.scirp.org/



DESCARGAR PDF




Documentos relacionados