From simulation to reality: a implementable self-organized collision avoidance algorithm for autonomous UAVs

Abstract

We present an implementable self-organized collision avoidance algorithm for UAVs (Unmanned Aerial Vehicle). The algorithm generates roundabouts to avoid collisions. It also calculates a safe reaction distance to start the collision avoidance maneuver based on relative velocity, UAV acceleration, communication frequency, and GPS inaccuracy. We also present our experimental testbed for autonomous UAVs, in which we have designed, developed, and validated the collision avoidance algorithm. The developed testbed comprises a modular architecture that includes a simulation environment, a set of real autonomous UAVs, and communication hardware. This architecture allows us to improve realistic models in the simulation and to test the same implementation in simulation as well as in real UAVs. We validate our collision algorithm in simulation and experiment and analyze in which conditions both results are similar and in which ones differ.