Search

This thesis presents several studies to improve and develop automated manufacturing systems along with introducing and evaluating the characteristics of various advanced multifunction composite structures. This research also provides high-performance composites that can bring various benefits for aerospace vehicles such as superior stiffness, lightweight, electrical conductivity, lightning strike protection, electrical grounding, and ductility that can be useful for frame integrity in crash load situations.

Robot grasping is emerging as an active area of research in robotics as the interest in human-robot interaction is gaining worldwide because of diverse industrial settings for sharing tasks and workplaces. It mainly focuses on the quality of generated grasps for object manipulation. However, despite advancements, these methods need to consider the human-robot collaboration settings where robots and humans will have to grasp the same objects concurrently. Therefore, generating robot grasps compatible with human preferences of simultaneously holding an object is necessary to ensure a safe and natural collaboration experience. In this work, we propose a novel, deep neural network-based method called CoGrasp that generates human-aware robot grasps by contextualizing human preference models of object grasping into the robot grasp selection process. We validate our approach against existing state-of-the-art robot grasping methods through simulated and real-robot experiments and user studies. In real robot experiments, our method achieves about 88% success rate in producing stable grasps that allow humans to interact and grasp objects simultaneously in a socially compliant manner. Furthermore, our user study with 10 independent participants indicated our approach enables a safe, natural, and socially aware human-robot objects' co-grasping experience compared to a standard robot grasping technique. To facilitate the grasping process, we also introduce a vision-language model that works as a pre-processing system before the grasping action takes place. In most settings, the robots are equipped with sensors that allow them to capture the scene, on which the vision model is used to do a detection task and objectify the visible contents in the environment. The language model is used to program the robot to make it possible for them to understand and execute the required sequence of tasks. Using the process of object detection, we build a set of object queries from the sensor image and allow the user to provide an input query for a task to be performed. We then perform a similarity score among these queries to localize the object that needs attention, and once identified, we can use a grasping process for the task at hand.

Increasing the manufacturing rate of carbon fibre structural components is a key goal for the aerospace industry. This thesis develops intelligent design optimization and manufacturing automation techniques for carbon fibre using machine learning.

In this research, a lower extremity exoskeleton was developed for people with mobility disorder. With the help of this device, individuals suffering from stroke or spinal cord injury or elderly people who have lost their muscle strength are able to stand up once again and move from one point to another. The device helps these individuals to regain their muscle strength after some time.

Three new monolithic compliant parallel micro/nano manipulators with different capabilities are introduced and the analytical modelling and computational analyses are presented. A novel optimisation approach is introduced utilising FEA for the structural optimisation of an entire 3D spatial micro/nano manipulator. Adaptive control methodologies are developed for high resolution and precise reference tracking of compliant manipulation systems. The experimental validation of the developed manipulation systems and control methodologies are conducted to present their behaviours including the in-depth dynamic and motion tracking performances, and hysteresis reduction. Comparative analyses are established to validate the obtained results from the experimentation and simulation.

In this paper we present the idea that by using AI planning in concert with formal task modeling, the overhead associated with plan creation for complex tasks can be reduced. The proposed approach uses a SysML taxonomy to model the system capabilities and the process specification, and the PDDL planning language to determine acceptable objective solutions. This idea is applied to the manufacturing domain, and examples are shown modeling a multi-robot system in an automobile manufacturing environment. A discussion is given regarding the merits of the demonstrated approach.

This thesis is concerned with the computer vision. More precisely, we are interested in the video real time tracking of 3D objects in video sequences, using only object appearances. An algorithm, developed earlier, allows to track textured planar patterns. It constitutes the core of our approach. We propose to adapt this 2D tracking method to the estimation of the 3D objects movement. For that, the 3D object is represented by a collection of reference images. For each one of these views, the 2D solution, quoted previously, allows to track fronto parallel movements (movement of the object in a plane parallel to the image plane). The aspect of the pattern representing the tracked object is not really modified by this motion. The delicate point, solved during this thesis, is the detection and the management of appearance changes of the tracked pattern due to relative rotations (between camera and object) in roll and pitch. In practice, the suggested approach allowed us the development of an experimental system dedicated to the tracking of human faces and the automatic navigation of a robotic arm, with a camera on its effector, around a 3D object., Ce travail de thèse s'inscrit dans le domaine de la vision artificielle. Plus précisément, nous nous sommes intéressés au suivi temps réel vidéo d'objets 3D mobiles dans des séquences d'images. A l'orgine de ces travaux, se trouve un algorithme, développé au LASMEA, dédié au suivi de motifs planaires texturés. Nous nous sommes proposés d'adapter cette méthode de suivi 2D à l'estimation du mouvement apparent d'objets 3D. Pour cela, l'objet 3D est modélisé à l'aide d'une collection d'images de référence. Pour chacune de ces vues, la solution 2D citée précédemment permet de suivre les mouvements fronto parallèles (déplacements de l'objet parallèlement au plan image) qui ne modifient pas de façon majeure l'aspect apparent dans l'image. Le point délicat, solutionné dans le cadre de cette thèse, est la détection et la gestion du changement d'aspect du motif suivi dû à des rotations relatives (caméra/objet) en site et azimut. Sur le plan pratique, l'approche proposée a permis le développement d'un système expérimental de suivi d'un visage et la navigation automatique d'un bras robotique, muni d'une caméra embarquée, autour d'un objet 3D.