Thèse Mosaic Réseaux Intelligents Multimodaux Basés sur l'IA pour la Surveillance Autonome des Moustiques Vecteurs dans un Contexte de Changement Global H/F Doctorat.Gouv.Fr

Établissement : Institut National Polytechnique de Toulouse École doctorale : GEETS - Génie Electrique Electronique,Télécommunications et Santé : du système au nanosystème Laboratoire de recherche : LAAS - Laboratoire d'Analyse et d'Architecture des Systèmes Direction de la thèse : Adam QUOTB ORCID 0000000177461364 Début de la thèse : 2027-09-01 Date limite de candidature : 2026-11-23T23:59:59 Le projet MOSAIC (Multimodal Smart AI Networks for Autonomous Mosquito Vector Surveillance under Global Change) vise à développer une nouvelle génération de systèmes intelligents et autonomes pour la surveillance des moustiques vecteurs de maladies dans un contexte de changement global. Les changements climatiques, les transformations des écosystèmes et l'évolution des activités humaines favorisent la modification de la distribution et de la dynamique des populations de moustiques, augmentant ainsi les risques d'émergence et de propagation de maladies telles que la dengue, le chikungunya, le Zika et le paludisme.

Le projet combinera imagerie, analyse acoustique, intelligence artificielle, capteurs intelligents et systèmes embarqués afin de permettre l'identification et le comptage automatisés des moustiques. La fusion de données multimodales permettra d'améliorer la fiabilité des systèmes de surveillance, notamment dans des environnements naturels complexes où les méthodes traditionnelles sont limitées.

La recherche sera structurée autour de quatre étapes : la constitution de bases de données multimodales, le développement d'algorithmes d'IA pour l'identification et le suivi des populations, l'intégration de ces algorithmes dans des plateformes autonomes à faible consommation énergétique, puis leur validation sur le terrain. Des campagnes expérimentales seront menées dans des environnements contrastés en France et en Équateur, notamment dans la forêt amazonienne à la Tiputini Biodiversity Station et aux Galápagos.

À terme, MOSAIC vise à fournir des outils de biosurveillance autonomes, en temps réel et déployables à grande échelle, capables de suivre les dynamiques spatiales et temporelles des populations de moustiques et de contribuer à la détection précoce de changements écologiques ou de risques vectoriels. Le projet s'inscrit ainsi à l'interface entre intelligence artificielle, écologie, systèmes embarqués, santé publique et approche One Health. Global changes are profoundly modifying ecosystems and increasing the risk of emergence and spread of mosquito-borne diseases worldwide. Climate change, biodiversity loss, land-use transformation, and increasing human mobility are altering mosquito habitats, geographical distributions, and seasonal dynamics, creating new conditions favorable to pathogen transmission [1-3]. Diseases such as dengue, chikungunya, Zika, and malaria are no longer restricted to tropical regions and are progressively becoming global health concerns [1,4]. The recent expansion of Aedes albopictus in Europe, including the occurrence of autochthonous transmission events in France, illustrates the urgent need for improved surveillance strategies [5,6]. The recent expansion of Aedes albopictus in Europe, including the occurrence of autochthonous transmission events in France, illustrates the urgent need for improved surveillance strategies [5,6]. Ecuador represents a particularly relevant study region for understanding the impacts of global changes on mosquito-borne diseases. Located at the intersection of highly contrasted ecosystems, from the Amazon rainforest to coastal and Andean regions, Ecuador exhibits strong environmental gradients that directly influence mosquito diversity, distribution, and pathogen transmission dynamics. The country is considered highly vulnerable to climate-driven changes in vector-borne disease risks, with recurrent outbreaks of dengue and the presence of multiple mosquito species of medical importance. The Amazonian ecosystem, in particular, provides a unique natural laboratory to investigate how biodiversity changes, environmental disturbances, and climate variability affect vector populations and disease emergence. Beyond public health impacts, changes in mosquito populations reflect broader ecosystem transformations, making vector monitoring a key component of global change assessment and One Health approaches [2,7].

Current mosquito surveillance strategies mainly rely on trapping campaigns followed by manual identification through morphological expertise or molecular methods. Although these approaches remain essential, they are limited by their cost, discontinuous sampling, requirement for specialized expertise, and inability to provide real-time information at large spatial scales [10]. Recent advances in artificial intelligence, computer vision, and acoustic analysis have opened new perspectives for automated insect monitoring [11,12]. Image-based approaches provide morphological and behavioral information, while acoustic signatures generated by wingbeats provide complementary information for mosquito recognition [13,14]. However, current approaches remain fragmented, often relying on single sensing modalities and laboratory-generated datasets. Their transfer to complex natural environments remains a major scientific challenge due to environmental variability, species diversity, acquisition conditions, and limited availability of representative datasets [11,15].


To overcome these limitations, this PhD proposes an integrated scientific strategy combining multimodal sensing, artificial intelligence, embedded systems, medical and ecological validation. The project is built on the hypothesis that the fusion of complementary information from images and acoustic signals will provide more robust and reliable mosquito identification than single-modality approaches, particularly under complex field conditions [11-14,16]. In parallel, the project hypothesizes that moving intelligence from centralized computing infrastructures to low-power embedded platforms will enable autonomous, real-time surveillance while significantly reducing communication requirements, latency, bandwidth usage, and energy consumption [17-19]. Following this rationale, the PhD will progressively develop multimodal datasets, design and optimize AI algorithms, integrate them into connected edge-computing sensing platforms capable of wireless data transmission, and validate the complete system through field deployments in France and Ecuador, including the Tiputini Biodiversity Station and the Galápagos Science Center. This stepwise strategy bridges the gap between laboratory-based proof-of-concept studies and operational biosurveillance systems capable of supporting One Health policies under global change scenarios.
The main objective of the MOSAIC project is to develop an autonomous, multimodal and intelligent system for the monitoring and identification of mosquito vectors under global change.

The specific objectives are to:

Develop multimodal sensing approaches combining images and acoustic signals for automated mosquito detection and identification.
Develop AI and machine-learning algorithms for mosquito classification, counting, and population monitoring.
Integrate AI into low-power embedded platforms capable of autonomous data acquisition, processing, and wireless transmission.
Validate the system under real field conditions in contrasting environments in France and Ecuador.
Develop ecological indicators and early-warning capabilities to monitor changes in mosquito populations, detect invasive species, and support vector-borne disease surveillance.
Demonstrate a scalable biosurveillance solution contributing to global change assessment and One Health strategies. Main previous achievements

The project builds upon the expertise and technological developments initiated within the PEPR- FOREST; MASSIF project, and the PEPS-CNRS, RCOP project recently obtained by Adam Quotb et al. , which addresses major challenges related to automated insect biodiversity monitoring. MASSIF demonstrated the importance of combining new trapping strategies, artificial intelligence, imaging technologies, acoustic sensing, and automated instrumentation to overcome current bottlenecks in large-scale entomological monitoring. The present project extends this scientific foundation toward a new application domain of major societal importance: the surveillance of mosquito vectors responsible for emerging infectious diseases. This transition from biodiversity monitoring to health-oriented biosurveillance represents a major scientific opportunity at the interface between ecology, artificial intelligence, embedded systems, and public health.

Beyond species detection and abundance estimation, the project will develop an integrated analytical and information system capable of extracting ecological indicators, monitoring the spatio-temporal dynamics of mosquito populations, and generating early-warning alerts when significant changes in vector abundance, distribution, or the establishment of invasive species are detected.

Scientific strategy

As described above, this project ambitions to develop, deploy and test standardized solutions to track and monitor mosquito disease vectors. The project will tackle all steps, from diagnostic tools to practical solutions, and from automated data acquisition to statistical analysis and real-time result visualization for both a scientific and a professional audience. Such a project would not have been feasible at large scale in the past, due to numerous barriers. In particular, sampling methods are so far not standardized among groups, and they necessitate regular field visits for trap collection. Moreover expert sorting and identification constitute a problematic bottleneck in monitoring even when specialists are available (which is often not the case). Our ambition is to bring together experts with complementary expertise from different disciplines and scientific cultures, providing the recruited PhD student with a unique interdisciplinary research environment and continuous access to leading expertise throughout the project. This collaborative framework will enable the project to deliver a viable solution for the scientific community, vector disease managers, and the general public to monitor mosquito diversity, analyze its spatial and temporal dynamics, and detect emerging threatening pests. The project will also benefit from the expertise of Biogents, a recognized actor in mosquito surveillance and trapping solutions, whose contribution will provide valuable knowledge on field deployment strategies, trapping optimization, and the requirements of operational vector monitoring. This interaction will facilitate the translation of technological developments into realistic surveillance solutions adapted to the needs of public health and environmental stakeholders.
To reach this goal and answer the Global Changes challenge of the BEST program, we gathered:
Medical Entomology experts, for an expert annotation of the databases developed in the project (photographs and sounds) and expert evaluation of the developed tools;
IA-specialists, both able to develop algorithms and to train and implement them;
smart sensors designers, to develop ad-hoc sensors and embedded IA solutions;
Experts in researching tropical ecosystems, with special interest in the Amazon region, Galapagos and France.
Expert in socio-economic stakeholders related to mosquito surveillance and trapping
Community and numerical ecologists, to propose relevant metrics and alert systems