Kermadj Islam

Internship Title: Holonomic Terrestrial Mobile Robot for 3D Mapping and Path Plannification Objective: The main goal of my internship is to enable the omnidirectional robot visualize accurately its 3D environement using a 3D LiDAR. Subsquentely, creating a projected 2D occupancy map that would be used later for path planning purposes, which takes into consideration the trajectory slops, robot dimensions, the mechanical constraints and the allowable motions of our robot through Nav2 Stack. Roles & Responsibilities: - Review all the works implemented on the omnidirectional robot platform through ROS2 - Analyzing literature by comparing the state-of-the-art approaches and the most adequate sensors to be implemented to create a 3D cartographer in order to visualize the robot's surroundings - Creating a 3D voxelized environement by implementing a 3D OctoMap using the 3D LiDAR - Generate a static TF2 between base_link and our Lidar to visualize the 3D map while the robot is moving - FIlteration of the floor and the robot's altitude to avoid projecting it later into a 2D map - Projection of the 3D filtered voxels into a 2D occupancy map and generate a global exploitable 2D map ready to use by the path planner - Calibrating the IMU in static mode and generating a dynamic TF2 between map and IMU to track its path - Enhancing the localization of the robot by implementing an Extended Kalman Filter, which fuses odometry data and IMU data in order to solve the issue of non-global optimization with OctoMap and the drift odometry caused by encoders. - Configurate the planner plugin SmacPlannerLattice as the planner_server in Nav2 - Integrate the projected 2D map as a Custom Costmap plugin in Nav2 Complementary work: - Maintenance of the robot by replacing the burned motor and enhance the globale wiring system - Solve the mechanical failure of the wheel axle with the motor - Helping a trainee colleague to implement a 2D semantic segmentation on ROS2

Thesis title: Specific person detection and identification using Deep Learning CNN for Real-Time drones’ applications. Objective: The main goal of this project was to make the drone detect and identify the target person that we are looking for in the crowds while the drone starts its aerial navigation. Subsequently, as soon as the target is identified, the drone starts the tracking process of the target. Roles & Responsibilities: - I was tasked for controlling the stability of a drone, made by a Bachelor student, with two different methods of control, Sliding mode and Backstepping Control, and then analyze the difference. - Preparing the software field and achieving the ideal combination of Robot Operating System ROS installed in Linux with the necessary drone packages and the appropriate version of OpenCV as well as the detection algorithm YOLO. - Implementation of an efficient and robust Multi-Layer Perceptron (MLP) for target tracking by drone. - Implementation of an algorithm for the detection and visual identification of the target using YOLO - Implementation of a 3D environment for the simulation of the system using Gazebo - Simulating and Real-time experiences in order to enhance the drone’s performances.

Roles & Responsibilities: - Wiring Electrical panel. - Supervision of electrical panel. - Creation of electrical diagram.

Roles & Responsibilities: - Testing and repairing electronic boards in the laboratory (oscilloscope, multimeter, welding) - Use of test benches for the diagnosis of on-board equipment failures - Analyzing aeronautical documentation (AMM, CMM, ATA chapters). - The workshops visited are: Electrical workshop, Electronics workshop, ATEC workshop, Radar and pulse workshop, Radio navigation workshop, Radio communication workshop, Battery workshop, Brakes and landing gear workshop.

Roles & Responsibilities: _ Study, analysis, and theoretical planning of the project, including the necessary calculations to prepare the optimal combination of electronic components, sensors, and actuators for the alarm system. -Designing and conception of the printed circuit board from A to Z using Sprint-Layout, followed by its integration into the alarm system. - Development of an efficient and robust algorithm for gas propagation detection, along with low-level programming of the sensors and actuators required for controlling the alarm system using the Arduino C language. - Integration of the alarm system with GSM to immediately notify the owner in case of gas leakage. - Simulation of the system's operation using Proteus, followed by real-time testing aimed at improving the performance of the alarm system.