Convergence of Multiagent Learning Systems for Traffic control