Computing optimal coalition structures in non-linear logistics domains

Abstract

We study computing optimal coalition structures in non-linear logistics domains, where coalition values are not known a priori and computing them is NP-hard. In our setting, the common goal of the agents is to minimise the system|s cost. Agents perform two steps: 1) deliberate appropriate coalitions; 2) exchange computed coalitions and generate coalition structures. We apply the concept of best coalition introduced in Sombattheera and Ghose (2008), to work in the non-linear logistics domains. We provide an algorithm and explain via examples to show how it works. Lastly, we show the empirical results of our algorithm in terms of elapsed time and number of coalition structures generated.