Addresses the problem of real-time planning and control of a group of aerial vehicles flying in an environment with obstacles and subject to disturbances and uncertainties. We present a novel methodology to the coordination of multiple heterogeneous vehicles that is based on a classical technique known as CL-RRT. We use a decentralized processing and communication to optimize the plan executed by each agent, providing a better solution to the execution step of the CL-RRT.
The technique is completely distributed and endows each agent with the capability to navigate through the environment while avoiding collisions with obstacles as well as with other team agents. The key is a prediction scheme executed by each robot that infers the behavior of the entire team at each time step. We show the applicability of the method in planning and executing waypoint navigation tasks for a group of MUAVs in outdoor environments in SE(3) and present real results.