Title: A cooperative learning approach to Mixed Performance Controller design: a behavioural viewpoint

Authors: Reyhaneh Bakhtiari, Babak N. Araabi, Majid Nili AhmadAbadi

Addresses: Control and Intelligent Processing Center of Excellence, Electrical and Computer Engineering Department, University of Tehran, Tehran 14395, Iran. ' Control and Intelligent Processing Center of Excellence, Department of Electrical and Computer Engineering, University of Tehran, and School of Cognitive Sciences Institute for Studies on Theoretical Physics and Mathematics, Tehran, Iran. ' Control and Intelligent Processing Center of Excellence, Department of Electrical and Computer Engineering, University of Tehran, and School of Cognitive Sciences Institute for Studies on Theoretical Physics and Mathematics, Tehran, Iran

Abstract: Based on the idea of cooperative learning in multiagent systems, two approaches to design Mixed Performance Controllers (MPCs) are presented. Each controller is assumed as an expert agent. These agents share their knowledge to construct a MPC with mixture of initial controllers| properties. In the first approach, the knowledge of each agent is extracted and presented in a Q-table – like tabular form. These tables are combined, and with greedy action – selection policy the MPC is found. In the second approach, controllers are combined from a behavioural viewpoint. Stability of the MPC is formalised in a theorem with a sufficient condition. Simulation studies indicate that the combined controller shows a mixture of desired properties associated with individual controller agents.

Keywords: cooperative learning; Q-learning; mixed performance controllers; MPCs; multiagent systems; multiobjective controllers; automatic learning; real time.

DOI: 10.1504/IJISTA.2007.012479

International Journal of Intelligent Systems Technologies and Applications, 2007 Vol.2 No.2/3, pp.137 - 160

Available online: 19 Feb 2007 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article