An Interconnected Dynamical System composed of dynamics-based Reinforcement Learning agents in a distributed environment: A case study

Author

Madera, Manuel ; Megherbi, D.B.

Author_Institution

Dept. of Electr. & Comput. Eng., Univ. of Massachusetts, Lowell, MA, USA

fYear

2012

fDate

2-4 July 2012

Firstpage

63

Lastpage

68

Abstract

This paper presents a case study of an Interconnected Dynamical System (IDS) composed of Intelligent Reinforcement Learning (RL) agents, and characterized by a Hybrid P2P/Master-Slave architecture. In particular, we propose and extent our previously proposed non-dynamics-based RL work to make it an IDS. Furthermore, we study how the addition of motion constrains, knowledge sharing between agents, and distributed computing affect the overall performance of the system. In addition, we introduce a new dynamics based reward mechanism for reinforcement learning agents.

Keywords

learning (artificial intelligence); peer-to-peer computing; IDS; RL; distributed computing; distributed environment:; dynamics-based reinforcement learning agents; hybrid P2P-master-slave architecture; interconnected dynamical system; nondynamics-based RL work; Equations; Heuristic algorithms; Learning; Mathematical model; Nickel; Turning; Vectors; Distributed multi-agent systems; Interconnected Dynamical Systems; Multi-agent systems; Reinforcement Learning; dynamics based reinforcement learning;

fLanguage

English

Publisher

ieee

Conference_Titel

Computational Intelligence for Measurement Systems and Applications (CIMSA), 2012 IEEE International Conference on

Conference_Location

Tianjin

ISSN

2159-1547

Print_ISBN

978-1-4577-1778-9

Type

conf

DOI

10.1109/CIMSA.2012.6269597

Filename

6269597