Democritus University of Thrace (DUTH)
Study and implementation of an innovative fully autonomous flying robotic system for life-saving services
Abstract
dc:descriptionThe major issue during crisis events occurrence is the immediate reaction and the provision of emergency services to humans in imminent danger. As human survival chances are negatively correlated with detection time, the response time is a vital factor in Search and Rescue (SAR) operations, thus potential delays may have dramatic and even lethal results. Moreover, the safety of rescue workers is another major issue and must be ensured in any circumstance. Unmanned Aerial Vehicles (UAVs) technology can provide a critical and multifaceted support in SAR missions. Given the fact that the faster these operations are planned and executed, the greater the chances of success they get, these aerial robots are ideal candidates for such applications owed to their ability of being nimble, quick-moving and easily programmed to exhibit autonomous behaviors. Hence, with the aim to increase the prospects of distressed people to remain alive, robotic rescue systems—either remote controlled or autonomous ones—must be instantaneously available, to provide a robust helping hand to rescuers, thus boosting their performance and ensuring their own safety. The doctoral dissertation in hand deals with the study and implementation of a novel fully autonomous UAV-based rescue platform for life-saving services, by combining Global Navigation Satellite System (GNSS) methods and state-of-the-art deep learning techniques. Specifically, the proposed Autonomous Aerial Rescue System (AARS), namely the ROLFER (RObotic Lifeguard For Emergency Rescue), is capable of precisely locating and providing instantaneous emergency services to humans in peril during critical events. The main system’s components are: a completely autonomous rescue UAV and its bearing equipment, the Ground Control Station (GCS) including all the apparent software and hardware for the system’s proper functionality, as well as the wearable equipment carried by humans in peril for the distress signal transmission. The rescue system’s Key Performance Indicators (KPI) are to: (i) be fully autonomous, i.e. guided solely by data provided by the distressed human’s wearable equipment without requiring to include human-in-the-loop, (ii) provide fast reaction as possible, i.e. to minimize the intervention time a UAV needs to provide invaluable assistance to a human in the context of an emergency and (iii) precisely detect the human in peril as well to accurately release the rescue equipment. The importance of not wasting the critical response time, as well as taking into serious consideration any other vital parameter during a SAR operation is obvious. The main objectives of this research are summarized as follows: The study, design and implementation of a fully autonomous aerial robotic system for the provision of life-saving services to humans in peril, The deployment of two human rescue scenarios in both maritime and terrestrial environments, The detection of human in peril using several accurate advanced Global Positioning System (GPS)/Global Navigation Satellite System (GNSS)-based localization methods, The detection, tracking and localization of human in peril using advanced image processing techniques as well as the precise rescue apparatus release, The hazard identification and risk analysis of the proposed AARS for the elimination of potential threats to both human in peril and rescue workers, by applying System Theoretic Process Analysis (STPA). Thus, the structure of the dissertation in hand is mainly based on the aforementioned objectives. In particular, Chapter 1 introduces the utilization of robotic systems in SAR operations including some crucial information about UAVs, a review of the related literature, as well the contribution of this dissertation within the state of art. Chapter 2 presents the proposed aerial rescue support system’s architecture and functionality, concerning details about both software and hardware configuration, as well as the system’s performance evaluation in terms of both accuracy and intervention time. Moreover, Chapter 3 addresses the issue of precise localization of the distressed human by implementing advanced geolocalization techniques (e.g. Real Time Kinematics (RTK)-GPS). Thus, an enhanced positioning system which grounds on a specially customized wearable device is proposed, allowing the precise positioning of human in peril. Both hardware and software configuration of the designed novel prototype are described. Experimental results from real flight tests of the proposed AARS and the effectiveness of the implemented moving-base RTK-GPS-based wearable apparatus are also discussed in detail. Chapter 4 presents the vision-based punctual and precise detection of humans in peril from the proposed AARS. In particular, human detection is implemented based on two different approaches:a)by utilizing deep learning techniques on-board the autonomous system’s rescue UAV for the human in peril in maritime environmentb)by combining RTK-GPS techniques and advanced image processing algorithms on the proposed rescue system’s GCS for the human in peril in terrestrial environment. Two suitable datasets, including human in peril in both maritime and terrestrial environments have been created for the computer vision algorithms training. Hardware and software configuration, as well experimental results for the accurate human detection and rescue apparatus fall down are presented in detail. In Chapter 5, a hazard identification and risk analysis of the proposed AARS, for the elimination of potential threats to both human in peril and rescue workers from the proposed system functionality is presented. The objectives of this analysis are a) to introduce the idea of analyzing the system by dividing it in separate operational modes and b) to feature the advantages of applying a systemic STPA-based safety approach on a SAR mission recruiting the use of UAVs. The dissertation is concluded in Chapter 6 where our findings are summarized and future work is mentioned.The doctoral dissertation in hand has been conducted in the Laboratory of Robotics and Automation of the Department of Production and Management of DUTh, which expertizes in such systems implementations.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Lygouras, Eleftherios
- Λυγούρας, Ελευθέριος
Subjects
dc:subject × 14- Μη-επανδρωμένα αεροχήματα
- Έρευνα και διάσωση εγκλωβισμένων ατόμων
- Υπολογιστική όραση και επεξεργασία εικόνας
- Βαθιά μάθηση
- Ανάλυση επικινδυνότητας
- Unmanned aerial vehicles (UAVs)
- Search and rescue
- Computer vision and image processing
- Deep learning
- Hazard analysis
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Μηχανολόγου Μηχανικού
- Engineering and Technology
- Mechanical Engineering
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/46958