Towards Human-UAV Physical Interaction and Fully Actuated Aerial Vehicles

Abstract

Unmanned Aerial Vehicles (UAVs) ability to reach places not accessible to humans or other robots and execute tasks makes them unique and is gaining a lot of research interest recently. Initially UAVs were used as surveying and data collection systems, but lately UAVs are also efficiently employed in aerial manipulation and interaction tasks. In recent times, UAV interaction with the environment has become a common scenario, where manipulators are mounted on top of such systems. Current applications has driven towards the direction of UAVs and humans coexisting and sharing the same workspace, leading to the emerging futuristic domain of Human-UAV physical interaction. In this dissertation, initially we addressed the delicate problem of external wrench estimation (force/torque) in aerial vehicles through a generalized-momenta based residual approach. To our advantage, this approach is executable during flight without any additional sensors. Thereafter, we proposed a novel architecture allowing humans to physically interact with a UAV through the employment of sensor-ring structure and the developed external wrench estimator. The methodologies and algorithms to distinguish forces and torques derived by physical interaction with a human from the disturbance wrenches (due to e.g., wind) are defined through an optimization problem. Furthermore, an admittance-impedance control strategy is employed to act on them differently. This new hardware/software architecture allows for the safe human-UAV physical interaction through exchange of forces. But at the same time, other limitations such as the inability to exchange torques due to the underactuation of quadrotors and the need for a robust controller become evident. In order to improve the robust performance of the UAV, we implemented an adaptive super twisting sliding mode controller that works efficiently against parameter uncertainties, unknown dynamics and external perturbations. Furthermore, we proposed and designed a novel fully actuated tilted propeller hexarotor UAV. We designed the exact feedback linearization controller and also optimized the tilt angles in order to minimize power consumption, thereby improving the flight time. This fully actuated hexarotor could reorient while hovering and perform 6DoF (Degrees of Freedom) trajectory tracking. Finally we put together the external wrench observer, interaction techniques, hardware design, software framework, the robust controller and the different methodologies into the novel development of Human-UAV physical interaction with fully actuated UAV. As this framework allows humans and UAVs to exchange forces as well as torques, we believe it will become the next generation platform for the aerial manipulation and human physical interaction with UAVs.