Centralized Collision-free Polynomial Trajectories and Goal Assignment for Aerial Swarms

Computationally tractable methods are developed for centralized goal\nassignment and planning of collision-free polynomial-in-time trajectories for\nsystems of multiple aerial robots. The method first assigns robots to goals to\nminimize total time-in-motion based on initial trajectories. By coupling the\nassignment and trajectory generation, the initial motion plans tend to require\nonly limited collision resolution. The plans are then refined by checking for\npotential collisions and resolving them using either start time delays or\naltitude assignment. Numerical experiments using both methods show significant\nreductions in the total time required for agents to arrive at goals with only\nmodest additional computational effort in comparison to state-of-the-art prior\nwork, enabling planning for thousands of agents.\n