This paper proposes a system partition model to improve the signal coordination control in large-scale urban arterials based on the classical MAXBAND methodology. The model divides the arterial system into several subsystems. Signal coordination is performed in each subsystem by considering the demand distribution along the whole arterial system. The obtained model is a mixed-integer nonlinear programming problem with multiple sets of binary variables. Numerical examples on an arterial of 20 signalized intersections show that the proposed model has potential to produce coordination plans that outperform those generated by Synchro.