COllective INtelligence with task assignment

In this paper we study the COllective INtelligence (COIN) framework of Wolpert et al. for dispersion games (Grenager, Powers and Shoham, 2002) and variants of the EL Farol Bar problem. These settings constitute difficult MAS problems where fine-grained coordination between the agents is required. We enhance the COIN framework to dramatically improve convergence results for MAS with a large number of agents. The increased convergence properties for the dispersion games are competitive with especially tailored strategies for solving dispersion games. The enhancements to the COIN framework proved to be essential to solve the more complex variants of the El Farol Bar-like problem.