Aerial Human-Comfortable Collision-free Navigation in Dense Environments

With current overuse of the road transportation system and planned increase in traffic, inno- vative solutions that overcome environmental and financial cost of the current system should be assessed. A promising idea is the use of the third dimension for personal transportation. Therefore, the European project myCopter, funded under the 7th framework, aimed at en- abling the technologies for Personal Aerial Transportation Systems as breakthrough in 21st century transportation systems. This project was the starting point of this thesis. When multiple vehicles share a common part of the sky, the biggest challenge is the man- agement of the risk of collision. While optimal collision-free navigation strategies have been proposed for autonomous robots, trajectories and accelerations for Personal Aerial Vehicles (PAVs) should also take into account human comfort for their passengers, which has rarely been the focus of these studies. Comfort of the trajectories is a key factor in order for this new transportation mean to be accepted and adopted by everyday users. Existing strategies used to maximize human-comfort of trajectories are based on path planning strategies, which compute beforehand the whole trajectory, implementing comfort as an optimization criteria. Personal Aerial Transportation Systems will have a high density of vehicles, where the time to react to potential threats might decrease to a few seconds only. This might be insufficient to compute a new trajectory each time using these path planning strategies. Therefore, in this thesis, a reactive decentralized strategy is proposed, maximizing the comfort of the trajectories for humans traveling in a Personal Aerial Vehicle. To prove the feasibility of collision avoidance strategies, it is not sufficient anymore to validate them only in simulation, but, in addition, real-time tests in a realistic outdoor environment should be performed. Nowadays, single drones can be effectively controlled by a single operator on the ground. The challenge relies instead on an efficient management of a whole swarm of drone. In this thesis, a framework to perform outdoor drone experiment was developed in order to validate the proposed collision avoidance strategy. On the one hand, an autopilot framework was developed, tailored for multi-drone experiments, allowing fast and easy deployment and maintenance of a swarm of drones. On the other hand, a ground control interface is proposed in order to monitor, control and maintain safety in a flight with a swarm of drones. Using the autopilot framework together with the ground control interface, the proposed collision avoidance strategy was validated using 10 quadrotors flying autonomously outdoor in a challenging scenario