Thèse Conception d'Un Essaim de Drones à Tolérance de Pannes pour l'Évaluation des Risques d'Inondations Soudaines H/F Doctorat.Gouv.Fr

Établissement : Ecole Nationale de l'Aviation Civile École doctorale : SYSTEMES Laboratoire de recherche : ENAC-LAB - Laboratoire de Recherche ENAC Direction de la thèse : Guthemberg DA SILVA SILVESTRE ORCID 0000000202737212 Début de la thèse : 2027-09-01 Date limite de candidature : 2026-11-23T23:59:59 Face au changement climatique rapide, le cycle de l'eau devrait être l'un des systèmes écologiques de la Terre les plus touchés. Par conséquent, une modélisation et une surveillance plus détaillées des cours d'eau sont essentielles pour évaluer avec précision le risque de catastrophes naturelles liées à l'eau de plus en plus fréquentes, telles que les crues soudaines provoquées par les précipitations. Dans ce contexte, plusieurs techniques de télédétection et cadres de modélisation ont été mis en oeuvre pour faciliter la surveillance environnementale et étudier l'impact continu du changement climatique d'origine humaine sur les cycles hydrologiques. De nombreux travaux de recherche s'appuient sur la télédétection par satellite pour offrir une perspective globale de la Terre. Cependant, les limites intrinsèques de résolution spatio-temporelle des plateformes satellitaires imposent des contraintes fondamentales à l'amélioration de la précision des systèmes de surveillance de l'eau. Pour pallier ces limites, le recours à des plateformes de surveillance moins coûteuses et plus flexibles, basées sur des essaims de drones volant à très basse altitude, s'est imposé comme une approche complémentaire prometteuse pour collecter des données à plus haute résolution grâce à des méthodologies d'échantillonnage plus souples. Notre projet de recherche doctoral commun vise à concevoir et à valider une plateforme de télédétection intelligente reposant sur un essaim de drones autonomes pour une surveillance précise des cours d'eau et une évaluation à la demande des risques de crues soudaines. One of the most promising emerging applications of drone swarms is the intelligent remote sensing of rapidly evolving natural phenomena that could lead to disasters. A typical example is a flash flood[10], especially in urban areas, where floods can result in many casualties and significant property damage. Unfortunately, such extreme events have become more frequent in recent years, possibly due to human-induced climate change. Therefore, airborne remote sensing using drone swarms can enable richer on-demand data collection and faster processing at the edge. This precise, high-resolution, real-time information on rapidly evolving and potentially dangerous rivers will help increase our understanding of these phenomena, leading to better prediction and, ultimately, better-informed decision-making during crisis response. While this technique shows great promise, experiments on concrete use cases are needed to better understand the advantages and limitations of drone swarms for hydrology and flood prediction.

Unlike currently used techniques based on in-situ data or satellite remote sensing, using drone swarms raises significant technical issues, including mission-level fault tolerance, autonomous configuration and coordination in wireless networks. Furthermore, the current legal frameworks that govern the use of drones for civil applications pose significant challenges for researchers and operators, particularly with regard to autonomous Beyond Visual Line of Sight (BVLOS) operations. For instance, the integration of civilian drone swarms into nationwide controlled airspace remains a work in progress, particularly in remote and isolated regions with inadequate networking infrastructure. For these reasons, using intelligent remote sensing with autonomous drone swarms remains challenging.

Iqbal et al. [6] identified several flood-management applications of drone technology, including flood inundation and flood-damage mapping, surface-water detection, flood boundary identification and flood victim identification. Drones are well-suited to disaster response due to their ability to provide near-real-time, high-resolution, spatially explicit information [7]. Images collected with drones have been processed using methods to estimate water extent, including convolutional neural networks [8] and supervised classification [9]. Jiménez-Jiménez et al. [14] emphasised the importance of having high-resolution images available for evaluating the impact of natural hazards. This data could be even more valuable to emergency response agencies if it were available within 24 hours of the event. Among other things, drones can identify affected structures, the extent of the damage and the number of impacted buildings and people [11].

Data obtained by drones can be used in hydrodynamic model simulations, including the identification of shorelines and the geometry of floodplains and the calculation of digital terrain models [12]. This allows the set-up and calibration of the hydrodynamic model, as well as updating and assessing its performance. Video images captured over a river and processed with the appropriate algorithms can estimate surface water velocity and consequently discharge [13]. This is another set of data that can support procedures related to hydrodynamic simulations.


[6] Umair Iqbal, Muhammad Zain Bin Riaz, Jiahong Zhao, Johan Barthelemy, and Pascal Perez. Drones for flood monitoring, mapping and detection: A bibliometric review. In: Drones 7.1 (2023), p. 32.

[7] Mbulisi Sibanda, Onisimo Mutanga, Vimbayi GP Chimonyo, Alistair D Clulow, Cletah Shoko, Dominic Mazvimavi, Timothy Dube, and Tafadzwanashe Mabhaudhi. Application of drone technologies in surface water resources monitoring and assessment: A systematic review of progress, challenges, and opportunities in the global south. In: Drones 5.3
(2021), p. 84.

[8] Asmamaw Gebrehiwot, Leila Hashemi-Beni, Gary Thompson, Parisa Kordjamshidi, and Thomas E Langan. Deep convolutional neural network for flood extent mapping using unmanned aerial vehicles data. In: Sensors 19.7 (2019), p. 1486.

[9] Przemysław Tymków, Grzegorz Jóźków, Agata Walicka, Mateusz Karpina, and Andrzej Borkowski. Identification of water body extent based on remote sensing data collected with unmanned aerial vehicle. In: Water 11.2 (2019), p. 338.

[10] Dottori, Francesco, et al. 'Cost-effective adaptation strategies to rising river flood risk in Europe.' Nature Climate Change 13.2 (2023): 196-202.

[11] Milan Erdelj, Enrico Natalizio, Kaushik R Chowdhury, and Ian F Akyildiz. Help from the sky: Leveraging UAVs for
disaster management. In: IEEE Pervasive Computing 16.1 (2017), pp. 24-32.

[12] Emilia Karamuz, Renata J Romanowicz, and Joanna Doroszkiewicz. The use of unmanned aerial vehicles in flood hazard assessment. In: Journal of Flood Risk Management 13.4 (2020), e12622.

[13] Paschalis Koutalakis, Ourania Tzoraki, and George Zaimes. UAVs for hydrologic scopes: Application of a low-cost UAV
to estimate surface water velocity by using three different image-based methods. In: Drones 3.1 (2019), p. 14.

[14] Sergio Iván Jiménez-Jiménez, Waldo Ojeda-Bustamante, Ronald Ernesto Ontiveros-Capurata, and Mariana de Jesús Marcial-Pablo. Rapid urban flood damage assessment using high resolution remote sensing data and an object-based approach. In: Geomatics, Natural Hazards and Risk 11.1 (2020), pp. 906-927. This joint doctoral research project aims to provide an innovative, drone-based solution to the increasingly frequent problem of flash flooding and the lack of precise spatial and temporal observation resolution for surveying it. The project will provide an opportunity to design novel modelling techniques and advanced sensor technologies for emerging, very high-resolution remote sensing applications on autonomous drone swarms. This includes open research problems in mobile edge computing, environmental monitoring, urban hydrological and climate modelling, and data collection for natural hazards such as flood risk prediction and assessment. Other areas of research include wireless communication, mission planning and unmanned aircraft system traffic management and regulations for autonomous drone swarms. The PhD project will be organized into four work packages covering the state-of-the-art research review, modelling, algorithmic, and experimental aspects of the research.

WP1 - Reporting on state-of-the-art remote sensing approaches and technologies (Months 1-8)
Objectives: Survey and report on the state-of-the-art research on remote sensing of river basins using drones swarms and flooding assessment techniques.
Expected outcome: A survey of intelligent drone swarms for flooding risk assessment and the description of promising case studies.

WP2 - Modelling river and its dynamical behaviour (Months 4-10)
Objectives: conduct numerical studies in order to prepare and improve the integration targeted application and case studies.
Expected outcome: a numerical model of the river and its dynamical behaviour - with coupled hydrological/hydraulic simulations of the response of the river to extreme rainfall events and floods development, and simulation of drone data and their use to survey/predict the river and flood risk in real time.

WP3 - Development of the on-demand autonomous remote sensing system for flash flooding (Months 6-24)
Objectives: To design, develop, and validate an innovative, cost-effective autonomous drone system capable of operating in swarms for the precise, efficient remote sensing and monitoring of river water levels, leveraging advanced technology and swarm intelligence to enhance environmental monitoring and data collection capabilities.
Expected outcome: Conceptual and system architecture design, formal specification and verification of system networked capabilities, sensor integration plan, swarm and flight planning behaviours, prototyping and software integration mostly through networks simulation and advanced emulation frameworks.

WP4 - Case study development, experimental analysis and concept validation (Months 18-36)
Objectives: Implement a proof-of-concept prototype for drone swarms to register the water level variation and the water velocity to calculate the discharge during the event and characterize properly the hydrography: rising, peak, falling, and base time.
Expected outcome: open source, fully reproducible prototype.