Year

2010

Degree Name

Doctor of Philosophy

Department

School of Computer Science and Software Engineering - Faculty of Informatics

Abstract

This thesis presents a number of algorithms for forming coalitions among cooperative agents in pragmatic domains where traditional cooperative game theory solution concepts do not apply due to bounded rationality of agents. While previous work in coalition formation in multi-agent systems research operated on relatively small number of agents, e.g. less than 30 agents, this work explores coalition formation among 100 agents, this is due to limited computational resources not the performance of the our algorithms. We explore a bestfirst search centralized algorithm for optimal coalition structures which is based on a novel idea of deciding what is the best coalition to put into coalition structure being generated. Empirical results show that the solution reaches optimality quickly and terminates quickly in pragmatic domains. We further explore on optimal coalition structures with distributed algorithms in linear and non-linear domains. For the linear domains, we explore linear production and integer programming. For the non-linear domains we explore logistic providers. Based on existing algorithms, we explore a novel environment of forming coalitions in supply networks involving buyers, sellers and logistics providers agents. In this setting, buyers form coalitions to increase their negotiation power while sellers and logistics providers form coalitions to aggregate their supply power and optimize their resources usage.

Download

02Whole.pdf (1208 kB)

Share

COinS
 

Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.