Robust Coalition Structure Generation

Tenda Okimoto (1) , Nicolas Schwind (2) , Emir Demirovic , Katsumi Inoue (3) , Pierre Marquis (4)

1 Kobe University
2 AIST - National Institute of Advanced Industrial Science and Technology
3 NII - National Institute of Informatics
4 CRIL - Centre de Recherche en Informatique de Lens

How to form effective coalitions is an important issue in multi-agent systems. Coalition Structure Generation (CSG) involves partitioning a set of agents into coalitions so that the social surplus (i.e. the sum of the rewards obtained by each coalition) is maximized. In many cases, one is interested in computing a partition of the set of agents which maximizes the social surplus, but is robust as well, which means that it is not required to recompute new coalitions if some agents break down. In this paper, the focus is laid on the Robust Coalition Structure Generation (RCSG) problem. A formal framework is defined and some decision and optimization problems for RCSG are pointed out. The computational complexity of RCSG is then identified. An algorithm for RCSG (called AmorCSG) is presented and evaluated on a number of benchmarks

Domaines: Intelligence artificielle [cs.AI]

Tenda Okimoto, Nicolas Schwind, Emir Demirovic, Katsumi Inoue, Pierre Marquis. Robust Coalition Structure Generation. 21st International Conference on Principles and Practice of Multi-Agent Systems (PRIMA'18), Oct 2018, Tokyo, Japan. pp.140-157. ⟨hal-03300961⟩