Rahim Kacimi

Widespread arthropod declines pose risks to ecosystem functioning and agriculture. Assessing this decline or potential remediation implies the need for standardized and scalable population monitoring. Image-based methods, including camera traps and citizen science programs, are increasingly used, but the volume of data collected requires automated analysis. Robust arthropod detection is essential for individual counting or fine-grained classification, yet current datasets and algorithms do not address the vast morphological diversity across arthropod species and often overlook the variety of photographic contexts, such as differences in background, lighting, and image composition, in which arthropods are captured. To address this gap, we developed an arthropod detection dataset, covering all terrestrial families present in France with available validated images on the iNaturalist platform (749 families). To achieve this, we employed an iterative workflow in which a YOLOv11 model pre-annotated images - using one representative species per family - followed by manual correction and model retraining. Repeating this process progressively reduced annotation effort and improved model accuracy. The final outcome consists of a publicly available curated detection dataset and a robust arthropod detector for natural background scenes. The detector achieves an F1-score of 0.91, demonstrating strong performance despite substantial interspecific morphological variation and heterogeneity in photographic contexts. We further demonstrated the taxonomical universality of the model showing high F1-score and IoU averaged at the class (0.79, 0.85) and order level (0.82, 0.86) and also a good detection generalizability (F1-score>0.90, IoU>0.83) on species, genera and families never encountered by the model during training. Finally, we show how this model can be improved to generalize to new datasets using data augmentation, complementary training data or fine-tuning and increase detection of small objects. In particular, we report performance of the improved models on three use cases largely used in non lethal insect monitoring: (i) diurnal pollinator monitoring through citizen science or (ii) flower and nocturnal insects monitoring through smartphone time-lapse of a UV-illuminated white panel. These results mark an important step toward automated analysis of arthropod images in natural contexts, from both large-scale automated monitoring approaches or from citizen science monitoring programs.

Les progrès technologiques réalisés ces dernières années ont permis le développement de nouveaux types de capteurs dotés de moyens de communication sans fil, peu onéreux et pouvant être configurés pour former des réseaux autonomes. Les domaines d'application sont nombreux : domotique, santé, domaine militaire ou bien encore surveillance de phénomènes environnementaux. Les limites imposées sont la limitation des capacités de traitement, de stockage et surtout d'énergie. La liberté laissée à l'implantation est forte et impose de concevoir complètement l'infrastructure, les mécanismes et les protocoles en fonction de l'application visée. Dans cette thèse, nous nous sommes tout d'abord focalisés sur des réseaux de petites tailles. Nous avons conçu une solution protocolaire "Placide" pour le suivi de la chaîne du froid proposée dans le cadre du projet ANR-CAPTEURS. L'originalité première de la solution repose sur l'absence d'infrastructure et de Station de Base. Elle est composée de protocoles performants fondés sur la formation d'un anneau virtuel entre les noeuds, auto-organisants et très économes en énergie. Le second axe est dédié à une étude expérimentale de la qualité du lien.L'objectif est double. Nous souhaitions tout d'abord étayer certaines hypothèses effectuées dans la première partie du travail. Il s'agissait ensuite de proposer des protocoles et des algorithmes fondés sur la qualité du lien. Nous nous sommes focalisés sur la variation de la qualité du lien en fonction de la distance entre les noeuds et de la puissance de transmission. Par la suite, l'impact de la qualité du lien sur la topologie du réseau a été étudiée. Les retours d'expérience sont importants quant à la compréhension des facteurs affectant la durée de vie du réseau. La dernière contribution concerne l'exploitation de ces observations en proposant des stratégies de partage decharge. Notre idée est que des protocoles très réduits et des mécanismes simples peuvent être mis en oeuvre pour le routage. Nous illustrons ces principes au travers d'exemples pour lesquels nous montrons la supériorité de ces solutions par rapport à des routages de type plus court chemin.

With the rapid expansion of Internet of Things (IoT) devices, current low-power wide-area network (LPWAN) technologies are increasingly susceptible to performance deterioration. LoRaWAN stands out as a leading LPWAN solution due to its ability to support long-range, low-power communication in massive IoT deployments. However, ensuring data reliability in dense and dynamic environments while optimizing transmission settings poses significant challenges, particularly in terms of interference mitigation and energy efficiency. To overcome these limitations, this article proposes an innovative approach that integrates K-means clustering and leverages reinforcement learning (RL) to effectively manage gateway placement and configure transmission parameters, such as spreading factors (SFs) and power levels, based on local signal conditions and environmental feedback. The model explicitly accounts for the capture effect and the imperfect orthogonality of SFs, allowing for more realistic and robust adaptation strategies without relying on the conventional centralized adaptive data rate (ADR) mechanism. The approach is validated through a real-world experimental deployment. Furthermore, we highlighted the novelty of our approach by a detailed analysis of LoRaWAN SF sensitivity thresholds and evaluated the impact of gateway densification and SF reallocation on network performance. Both experimental and simulation outcomes show significant improvements in data extraction rate (DER) and energy efficiency, particularly in large-scale, congested network scenarios.

Locating at-risk workers in hospitals using legacy private 5G networks is a daunting task that involves solving the problem of indoor localization using commercial off-the-shelf proprietary hardware. Currently, no full-stack schemes or realistic indoor positioning experiments have been conducted using 5G. In this study, we present the first comprehensive 5G framework that combines fingerprinting with the 3GPP Enhanced Cell ID (E-CID) approach. Our methodology consists of a machine-learning model to deduce the user's position by comparing the signal strength received from the User Equipment (UE) with a reference radio power map. This challenging method has four main contributions. First, the 3GPP protocols and functions are extended to provide open, secure, and universal core network-based localization functions. Second, to generate a reference map, the first paradigm of Optical Radio Power Estimation using Light Analysis (ORPELA) is introduced. Real-world experiments prove that it is reproducible and more accurate than state-of-the-art radio-planning software. Third, machine-learning models are designed, trained, and optimized for an ultra-challenging radio context. Finally, an extensive experimental campaign is conducted to demonstrate the expected indoor localization performance of realistic 5G private networks.

Although Unmanned Aerial Vehicles (UAVs)-aided wireless sensor networks (WSNs) have gained many applications, it is not for long that research works have been produced to define effective algorithms and protocols. In this article, we address the UAV-enabled WSN (U-WSN), explore the performance and the capability of the UAV, define the UAV functionalities as a communication node, and describe the architectures and the relevant typical technologies that emerge from this new paradigm. Furthermore, this article also identifies the main factors which influence the U-WSN design and analyzes the open issues and challenges in U-WSN. These insights may serve as motivations and guidelines for future designs of UAV-enabled WSNs.