On Autonomous Deployment of Micro Aerial Field Vehicles

In this thesis, I am going to investigate the control, navigation and path planning frame-works forMicro Aerial Vehicles (MAVs), both mathematically and algorithmically. Inorder to deploy them in challenging real life applications and with a main focus on theunderground mine navigation and wind turbine inspection. As it will be presented, theobjective of the proposed modules is to provide to the robotics community a foundationon field robotics, which one can continue research on.Towards this envisioned aim, this thesis will establish the following main theoreticaland practical contributions: 1) the unifiedNonlinear Model Predictive Control (NMPC)framework for position-velocity reference tracking, while guaranteeing collision free pathsby considering obstacles. In this framework, theNMPCconsiders also the localizationuncertainties and tracks either position or velocity references, or none of them if neces-sary. The obstacles are extracted from 2D/3D point clouds and in the sequel are definedin local coordinates that result to their decoupling from localization uncertainties. Thus,the obstacle avoidance remains active and guarantees collision avoidance in all the cases.The proposed control architecture allows for a mission continuation by aMAVevenin the presence of a localization drift, providing recovery opportunities to the localiza-tion scheme in case of loop closure events or reaching feature-rich areas. 2) Introducinga baseline approach for the deployment of aerial underground scouts for subterraneanenvironments. The proposed framework focuses on the navigation, control and visioncapabilities of the aerial platforms with low-cost sensor suites, contributing significantlytowards real-life applications. The thesis proposes multiple methods for correcting theheading of theMAVtowards open spaces in featureless dark tunnel environments forfast fully autonomous navigation. 3) Developing a holistic approach to the problems of2D area coverage withMAVs for polygon areas, while considering the camera footprint.In the presented novel approach a 3Degree of Freedom (DoF)camera movement is con-sidered and the shortest path from the taking off to the landing station is generated,while covering the target area. 4) Finally, this thesis revisits theCollaborative CoveragePath Planning (C-CPP)problem for the inspection of complex infrastructures, and es-tablishes a theoretical framework, capable of offline providing a path for accomplishinga full coverage of the infrastructure with multipleMAVs