When tracking a target in a sensor network with constrained resources, the target state estimate error can be significantly reduced using non-myopic sensor scheduling strategies. Integer non-linear programming has been used to obtain myopic sensor schedules . In this paper, we apply it to a non-myopic sensor scheduling scenario consisting of a network of acoustic sensors in a centralized sensor network; there is one fusion center that combines measurements to update target belief. We cast this problem, which we call the Central Node Scheduling problem, as an integer non-linear programming problem with the objective of minimizing the total predicted tracking error over an M step planning horizon subject to sensor usage and start-up cost constraints. Using Monte Carlo simulations, we show the benefits of this approach for the centralized sensor network.