Perception Aware Guidance Framework for Micro Aerial Vehicles

Sammanfattning: Micro Aerial Vehicles (MAVs) are platforms that have received significant research resources within robotics community, since they are characterized by simple mechanical design and versatile movement. These platforms possess capabilities that are suitable for complex task execution, in situations which are impossible or dangerous for the human operator to perform, as well as to reduce the operating costs and increase the overall efficiency of the operation. Until now they have been integrated in the photography-filming industry, but more and more efforts are directed towards remote reconnaissance and inspection applications. Moreover, instead of carrying only sensors these platforms could be endowed with lightweight dexterous robotic arms expanding their operational workspace allowing active interaction with the environment, capabilities that can be vital for applications like payload transportation and infrastructure maintenance. The main objective of this thesis is to establish the concept of the resource-constraint aerial robotic scout and present perception aware frameworks for guidance of the platform and the aerial manipulator as part of the enabling technology towards fully autonomous capabilities. The majority of the works has been developed aiming the application scenario of the MAV deployments in subterranean environments for search and rescue missions, infrastructure inspection and other tasks. A key factor when deploying aerial platforms in dark and cluttered underground tunnels in the lack of illumination which degrades the performance of the visual sensor. It is essential for the inspection or reconnaissance task to get visual feedback from the robot and therefore, this thesis evaluates methods for low light image enhancement in real environments and with datasets collected from flying vehicles, while proposes a preprocessing methodology of the visual dataset for enhancing the 3D mapping of the area. Another capability required when deploying the platforms is the navigation along the tunnel. This thesis establishes robocentric Non Linear Model Predictive Control (NMPC) framework for fast fully autonomous navigation of quadrotors in featureless dark tunnel environments. Additionally, this work leverages the processing of a single camera to generate direction commands along the tunnel axis, while regulating the platform’s altitude. Finally, combining the agility of MAVs with the dexterity of robotic arms leads to a new era of Aerial Robotic Workers (ARWs) with advanced capabilities, suitable for complex task execution. This technology has the potential to revolutionize infrastructure maintenance tasks. The development of efficient and reliable perception modules to guide the aerial platform at the desired target areas and perform the respective manipulation tasks is, among others, an essential step towards the envisioned goal. Thus, the aim of this work is the establishment of a visual guidance system to assist the aerial platform before applying any physical interaction. The proposed system is structured around a robust object tracker and is characterized by stereo vision capabilities for target position extraction, towards an autonomous aerial robotic worker.