Adnane Addaim

In the context of intelligent transportation systems (ITS) development, vehicle-to-everything (V2X) communication plays a central role by enabling information exchange between vehicles (V2V), infrastructure (V2I), pedestrians (V2P), and the network (V2N). The effectiveness of these systems relies heavily on the performance of the antennas employed, which must meet strict requirements in terms of compactness, bandwidth, gain, and electromagnetic compatibility. One of the main challenges lies in designing antennas suitable for the embedded vehicular environment, where space is limited and the propagation conditions are complex. In this context, the present study aims to design, simulate, fabricate, and experimentally evaluate a planar inverted-F antenna (PIFA) dedicated to V2X communication in the 5.8 GHz band. The primary objective is to develop an antenna that is both compact and high-performing, tailored to the specific constraints of V2X applications. The adopted methodology involves a comprehensive parametric study, focusing on several key design parameters that influence the antenna’s performance, such as substrate selection, feeding point location, and the addition of a slot in the structure. These factors are analyzed to optimize the radiation characteristics, resonant frequency, and impedance matching of the antenna. The results demonstrate the feasibility of a PIFA antenna that offers an excellent trade-off between miniaturization and performance, making it well suited for V2X communication applications at 5.8 GHz.

Deep learning has been increasingly utilized for cloud detection in Earth observation nanosatellites, offering effective solutions to enhance mission performance. Traditional methods of image capture, onboard storage, and transmission face challenges such as bandwidth limitations and the transmission of cloud-obstructed images, highlighting the need for efficient onboard artificial intelligence. However, energy consumption remains a critical consideration for onboard processing, particularly in nanosatellites where resource constraints are significant. To address these challenges, we propose an optimized SegNet-based deep learning model implemented on the STM32H7 microcontroller for real-time cloud detection, designed to operate within the nanosatellite's strict energy budget. This work, conducted within the project of the UM5-EOSat nanosatellite mission, utilized captured images from the Gecko imager for model evaluation. The customized SegNet architecture, tailored with minimal kernels and layers, achieved an accuracy of 93.50%, effectively balancing performance and computational efficiency. Quantization further optimized energy consumption, achieving a reduction of 82.2% at 280 MHz. The quantized model demonstrated a memory footprint of 304 KB RAM and 110 KB Flash memory, with an inference time of 0.21 s and an energy consumption of 31.41 mJ, ensuring compatibility with the nanosatellite's resource constraints.

In recent years, there has been growing interest in leveraging the Internet of Things (IoT) and Artificial Intelligence (AI) technologies for agriculture. A significant challenge for developers in this field is creating applications that provide precise data about plants, facilitating the smart automation of plant management. This paper presents Plantonome, an open-source application developed using the Flutter software development kit (SDK) and the Dart programming language. Designed to integrate with IoT devices, Plantonome quickly and accurately identifies ornamental plant genera or species using the Plant.id API for plant image analysis. The application also utilizes a NoSQL database for storing user data and plant preferences, and it includes a dataset of ornamental plants with details such as name, brightness, temperature, and humidity requirements. The development approach outlined in this paper accelerates the creation process and results in a high-performing application with a flexible user interface and smooth user experience. The application, tested on Android 5.0 (API level 21) or higher, achieved an accuracy of 94.64% for plant identification and received highly positive feedback regarding its functionality, usability, and efficiency. This work offers significant benefits to researchers and startups aiming to develop cross-platform applications that can automate various agricultural tasks, contributing to advancements in smart agriculture.

Recently, constellation of satellites has drawn a lot of interest from academia and industry as potential solution for extensive coverage of wide range of internet of things (IoT). In this work, IoT devices was assumed to be covered by constellation of low-earth orbit (LEO) satellites where the medium access control (MAC) technique called slotted ALOHA is employed. In this article, we use a constellation of satellites to reduce the collision domain and enhancing performance in order to obtain maximum results. We have carried out some modeling and simulations to optimize the number of satellites with different erasure probabilities with respect to IoT devices in order to enhance throughput and stability of slotted ALOHA protocol using the network simulator 2 (NS2). The numerical results have shown an improvement in terms of throughput and stability. And the simulation of the same system using NS2 is conducted and shows a good correlation with the theoretical study. Where the throughput reached 0.82% instead of 0.52%. Our findings offer proof that this method helps to use large number of IoT, and reduce collisions compared to conventional slotted ALOHA.

In order to improve the selection of multi-point relays (MPRs) by a node node performing the computation (NPC) in the optimized link state routing (OLSR) protocol and therefore to guarantee more security for the routing in the mobile ad hoc network (MANET), we propose new approach that could distinguish between the strong and weak MPRs in the list of MPRs already selected using the standard algorithm described in RFC3626 document. This approach is based on self organizing map (SOM) artificial neural network that processes the collected data and then only selects the strong MPRs using a set of criteria allowing a reliable retransmission and a strong link and therefore better network performances. The obtained results, from the simulations that have been carried out using a customized network simulator 3 (NS3) network simulator, show an improvement in terms of throughput, packets delivery ratio (PDR) and the security of the network compared to the standard approach.