We present a real-time implementation of a distributed motion planning framework that is based on model predictive control with one step prediction horizon and submodular function minimization. In particular, our focus is to evaluate the real-time performance of this distributed motion coordination framework. For performance evaluation, we develop a realistic simulation environment for the challenging setup of capture the flag game, which is played between two teams. We consider a scenario in which each team has four quadcopters and the game is played in an arena with multiple obstacles. We develop the simulation setup primarily in Gazebo with software in the loop. The software in the loop is the autopilot software, which is used to stabilize and control the motion of each quadcopter. The motion plan for the defense team is computed by minimizing submodular potential functions using the distributed and online algorithm presented in our previous work. Based on extensive simulations under various conditions, we verify that the proposed approach can be used effectively for real-time distributed control of multiagent systems over discrete input space.