Physical interaction, aerial construction, and coordinated flight for flying machines

This thesis investigates flying machines interacting with their environment and the design of coordinated multi-vehicle flight. Specifically, it presents control strategies and algorithms that allow for quadrocopters to physically interact with people, to autonomously realize load-bearing structures, and to easily coordinate the motion of multiple vehicles. Each of the algorithms presented in this thesis is experimentally verified in the Flying Machine Arena, a testbed for aerial robotic research. A control scheme based on admittance control is used for physical human-quadrocopter interaction. The control scheme consists of a model-based estimator, which estimates the external forces and torques acting on the vehicle using position and attitude information, an underlying position and attitude controller, and an admittance controller. The closed-loop characteristics are investigated in detail for the near-hover case and an interaction detection strategy allows for a reliable user interaction. The ability of flying machines for aerial construction is explored in two ways. On the one hand, the assembly of a 6-m-tall tower by quadrocopters acts as a proof-of-concept, demonstrating the ability of aerial vehicles to transport and place construction elements to otherwise inaccessible locations. Reliable strategies to pick up the foam modules and place them at the appropriate locations are developed. Furthermore, the process and challenges of deploying such a complex system for a live exhibition are discussed. On the other hand, the thesis develops a control architecture and hardware solutions that allow for flying machines to deploy ropes and assemble load-bearing tensile structures. Taking into account the effects of the rope on the flying machines, techniques to realize various type of building elements, such as links, nodes, and braids are investigated. A 7.4-m-long rope bridge acts as a demonstrator, showing for the first time that small flying machines are capable of autonomously realizing load-bearing structures at full-scale. This thesis also addresses the design and automatic generation of coordinated trajectories for a fleet of flying machines. It presents an algorithm to generate three-dimensional collision-free trajectories for multiple flying machines. The problem is solved within seconds using sequential convex programming, where convex approximations are used for non-convex constraints. The algorithm allows to generate feasible trajectories from a set of initial states to a set of final states, while ensuring that a minimum distance between vehicles is kept. These results are integrated into a software tool that enables the design of aerial dance performances.