A Local Adaptive Perfect Coalition Structure Generation Problem Considering Coalition Breakup

Forming effective coalitions is a major issue in multiagent systems. The existing coalition structure generation (CSG) algorithms fail to determine the structure ´s robustness because agent withdrawals cannot be identified in all the coalition structures. The coalition lattice, which was proposed in a previous study, provides an evaluation of the coalition structure´s robustness and enables to apply the existing CSG algorithms. However, to construct the coalition lattice, the lower bounds for the coalition utilities must be considered. In the present study, robustness is redefined and an improved conversion algorithm is proposed. In particular, a method to estimate the upper and lower bound of the coalition values is presented. Furthermore, an effective algorithm to convert the problem of finding the coalition structures´ robustness into a CSG problem is proposed. A case-study evaluation of the proposed method shows that it constructs the coalition lattice more efficiently.