A robust real-time path planner for the collision-free navigation of multirotor aerial robots in dynamic environments

The development of deliberative capabilities is required to achieve an intelligent fully autonomous behavior of unmanned aerial systems. An important deliberative capability is the generation of collision-free paths in complex environments. This paper presents a robust real-time collision-free path planner used for the horizontal 2D navigation of multirotor aerial robots in dynamic environments. Its design, using geometric primitives to describe the environment combined with a launching time generation of a probabilistic roadmap graph, permits an efficient management of dynamic obstacles. The use of an A∗ discrete search algorithm, together with a potential field map as the cost function, allows to speed up the collisionfree path computation ensuring that it never falls in local minima. Additionally, the velocity and acceleration along the collision-free planned path is calculated. The performance of the proposed path planner is evaluated in this paper with two simulations with complex environments including a labyrinth and dead ends, and with a real flight experiment where three fully autonomous aerial robots executed an emulated search and rescue mission. The proposed path planner has been released to the scientific community as an open-source software included in Aerostack[1]. In addition, it has extensively been used in multiple research projects with real flights, demonstrating its good performance.