A Cooperative Approach for Autonomous Landing of UAVs

• Autor(a) principal: Prates, Pedro Alexandre de Sousa
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/51002
• topic: UAV landing systems; Robotic cooperation; Background Subtraction; Optical Flow; motion patterns; high-level control; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

This dissertation presents a cooperative approach for the autonomous landing of MRVTOL UAVs (Multi Rotor-Vertical Take-off and Landing Unmanned Aerial Vehicles). Most standard UAV autonomous landing systems take an approach, where the UAV detects a pre-set pattern on the landing zone, establishes relative positions and uses them to perform the landing. These methods present some drawbacks such as all of the processing being performed by the UAV itself, requiring high computational power from it. An additional problem arises from the fact most of these methods are only reliable when the UAV is already at relatively low altitudes since the pattern’s features have to be clearly visible from the UAV’s camera. The method presented throughout this dissertation relies on an RGB camera, placed in the landing zone pointing upwards towards the sky. Due to the fact, the sky is a fairly stagnant and uniform environment the unique motion patterns the UAV displays can be singled out and analysed using Background Subtraction and Optical Flow techniques. A terrestrial or surface robotic system can then analyse the images in real-time and relay commands to the UAV. The result is a model-free method, i.e independent of the UAV’s morphological aspect or pre-determined patterns, capable of aiding the UAV during the landing manoeuvre. The approach is reliable enough to be used as a stand-alone method, or be used along traditional methods achieving a more robust system. Experimental results obtained from a dataset encompassing 23 diverse videos showed the ability of the computer vision algorithm to perform the detection of the UAV in 93,44% of the 44557 evaluated frames with a tracking error of 6.6%. A high-level control system that employs the concept of an approach zone to the helipad was also developed. Within the zone every possible three-dimensional position corresponds to a velocity command for the UAV, with a given orientation and magnitude. The control system was tested in a simulated environment and it proved to be effective in performing the landing of the UAV within 13 cm from the goal.