Bio-inspired Implicit Communication and Control for Quadcopter Drone Fault-Tolerant Formation Flight.

The first main objective of this Thesis research project is to develop a drone following algorithm based on the study of drones' interaction signals. This algorithm must be able to run on drones with low computational power as the ones used to develop it. The idea behind this algorithm stands in the geese' formation flight, where such birds are able to fly in an optimal wing vortex point to save energy. This results in their typical V-shaped swarm formations. Similarly, the used drones must be able to detect the wind produced by other's rotors to engage a following procedure. The work starts from a previous thesis research of Paolo Ceppi and aims to continue his work to achieve a new goal. This research work is divided into 3 main parts as 3 main actions were completed. At first, the drone interaction model had to be built by performing some real-world tests and by analyzing the results. After that, a solution was found to the problem using a multiple detection approach: the drone that follows would approach the leader multiple times and register the necessary information to estimate its speed. This algorithm also takes into account all cases of wrong speed detection by periodically forcing a new leader search. Later, it was noticed that the drone-following algorithm could be used to actually transmit data from the leader to the follower; in fact, at any point, the leader could perform a specific speed pattern that will be interpreted by the follower as a command. Here there is the second part of this thesis title: this functionality can be used in case a direct communication channel between drones fails. A working leader, in that case, would still be able to give commands to the faulty elements, such as "go back home" or "land". Finally, it was noticed that the available drones weren't able to fly closely as their ultrasonic sensors' beam angles were big enough to cause altitude control disturbances. As a solution to this problem, a new altitude estimation algorithm was designed exploiting available elements in the system, such as the camera and rotors' power output. This algorithm consisted of completely redesigning and optimizing the Image processing system for a better data rate and also redesigning the state estimator block. The obtained result exploits the image signal as validation and the thrust signal as instant values because of its higher data rate. However, it was demonstrated that using a different platform with optical sensors might be a much easier solution (less computationally expensive).