Hierarchical, Multi-layer and Human Aware Solution For The Deployment Of Autonomous Aerial Robots In Large Areas Monitoring And Inspection Scenarios

The use of robotics and autonomous systems in various industries and everyday life has resulted in a need for engineering solutions that can simplify the interaction between humans and robots to reduce physical and cognitive workload, particularly during complex or haz- ardous tasks. Despite the increasing popularity of flying robots as valuable tools in Industry 4.0, there is still a research gap in the autonomous monitoring and inspection of large areas or infrastructures. This gap also exists in search and rescue missions, especially in exploring indoor and constrained environments, where aerial robotics agents are still challenging to manage, requiring an expert pilot in the rescue team. To address these challenges, this thesis proposes an innovative hierarchical multi-layer scalable solution for the employment of an aerial robot fleet in activities on a large area scenario, based on a strategic and tactical level. The thesis introduces a novel concept of a human-aware drone activity planner system called Social Drone Sharing (SDS), which is a multilayer system based on the cloud that integrates the communication between authority and resident volunteers with the purpose to organize and manage a fleet of flying robots on a large monitor scenario. At the higher strategic level, innovative routing and path optimization solutions based on Linear Programming are proposed, which enable the drones to navigate safely between some important locations that need to be visited, and the social charging stations, defined as landing spots provided by resident volunteers who give their availability to land the drone in their private area. At a lower tactical level, the thesis provides insight into autonomous robot applications in outdoor and indoor monitoring and inspections scenarios, with particular attention to the human interaction and integration component. An innovative computer visual servoing algorithm based on image segmentation and multi-sensors fusion techniques for inspections of large infrastructure is introduced. The reliability of the system has been studied in real inspection experiments conducted on a Photovoltaic Power plant, where the robot was able to track efficiently multiple PV lines, increasing the quality of the collected data, and decreasing human effort. To increase the v efficiency of the tracking, an Extended Kalman Filter fuses multiple observations coming from a thermal and an RGB camera. Search and rescue applications or monitoring applications can be problematic in complex, indoor, and constrained environments, especially when drones are involved, requiring an expert pilot for safe flight. To increase the level of autonomy of the drone during these tasks presented at a lower tactical level, a tele-immersive human-robot cognitive and phys- ical collaborative interaction framework through Mixed Reality has been proposed. This framework increases the role of the drone as an effective teammate since the user and the robot perception are now shared in a complete innovative solution. Furthermore, a Variable Admittance Control (VAC) coupled with a planning algorithm enables the user to interact with the robot using a gesture-based interaction. The planning algorithm acts on the drone perceived environments and drives the user during the exploration along an obstacle-aware path. In summary, this thesis proposes solutions for monitoring large and dislocated areas, connecting people with authorities, and conducting autonomous and vision-based inspections of large infrastructure. The proposed approaches can also be applied to Human-Drone Interaction, especially when the robot needs to collaborate with humans in indoor constrained environments.