Systems engineering introduces the notion of top-down design, which involves viewing an entire system comprised of its components as a whole functioning unit. This requires an understanding of how those components efficiently interact, with optimization of the process emphasized rather than solely focusing on micro-level system components. The traditional approach to the systems engineering process involves requirements decomposition and flow down across a hierarchy of decision making levels, in which needs and requirements at one level are transformed into a set of system product and process descriptions for the next lower level. This top-down requirements flow approach therefore requires an iterative process between adjacent levels to verify that the design solution satisfies the requirements, with no direct flow between nonadjacent hierarchy levels. This thesis introduces a methodology that enables decision makers anywhere across a system-of-systems hierarchy to rapidly and simultaneously manipulate the design space, however complex. A hierarchical decision making process will be developed in which a system-of-systems, or multiple operationally and managerially independent systems, interact to affect a series of top level metrics. This takes the notion of top-down requirements flow one step further to allow for simultaneous bottom-up and top-down design, enabled by the use of neural network surrogate models to represent the complex design space. Using a proof-of-concept case study of employing a guided projectile for mortar interception, this process will show how the iterative steps that are usually required when dealing with flowing requirements from one level to the next lower in the systems engineering process are eliminated, allowing for direct manipulation across nonadjacent levels in the hierarchy. For this system-of-systems environment comprised of a Monte Carlo based design space exploration employing rapid neural network surrogate models, both bottom-up and top-down design analysis may be executed simultaneously. This process enables any response to be treated as an independent variable, meaning that information can flow in either direction within the hierarchy.

Ce mémoire de thèse traite du développement des dynamiques et des lois de commande de vol d’un projectile d’artillerie gyrostabilisé guidé par une tête découplée. Un modèle non linéaire du projectile est proposé, et sert à calculer un modèle linéarisé de la dynamique de roulis du nez et un modèle q-LPV des chaînes de tangage et de lacet à paramètres fortement variants. Les incertitudes de modélisation sont prises en compte pour concevoir l’autopilote. Des propriétés importantes des projectiles gyrostabilisés, qui sont liées au couplage dynamique tangage/lacet, aux modes internes et à la stabilité, sont mises en valeur grâce au modèle q-LPV. En vue de l’utiliser pour calculer une loi de commande, la dimension de son vecteur des paramètres est réduite et la position des capteurs intégrés dans le nez est considérée. Un seul correcteur linéaire est suffisant pour la dynamique de l’angle de roulis du nez alors qu’une stratégie systématique de commande par séquencement de gains basée sur une linéarisation est élaborée séparément pour générer un correcteur séquencé des facteurs de charge de tangage et de lacet. Des structures de commande fixées d’ordre réduit sont conçues en appliquant la même approche de synthèse linéaire H∞ par façonnage de gain de boucle pour les axes de roulis et de tangage/lacet. De très bonnes propriétés de performance et de robustesse en boucle fermée, comparables à celles fournies par des correcteurs d’ordre plein, sont obtenues. Finalement, l’efficacité de l’autopilote augmenté d’une loi de guidage par navigation proportionnelle pure est vérifiée via de nombreuses simulations non linéaires de trajectoires. Ces dernières correspondent à des scénarios de vol nominaux d’interception de cibles balistiques, non balistiques immobiles, ou manœuvrantes, ainsi qu’à des scénarios considérant des perturbations sur les conditions de tir ou sur les dynamiques du projectile guidé / This thesis addresses the development of the flight dynamics and control laws for an artillery spin-stabilized projectile equipped with a decoupled guidance nose. A projectile nonlinear model is discussed, and it is used for computing a linearized model of the nose roll dynamics along with a q-LPV model of the highly parameter-varying pitch/yaw-dynamics. Modeling uncertainty is taken into account for autopilot design. Important properties specific to spin-stabilized projectiles, which are relevant to pitch/yaw-channel cross-coupling, internal modes and stability, are highlighted using the q-LPV model. In order to use the latter for calculating a control law, the dimension of its parameter vector is reduced and the position of the nose-embedded sensors is considered. A single linear controller is sufficient for the nose roll angle dynamics whereas a systematic linearization-based gain-scheduled control strategy is separately devised to provide a pitch/yaw-axis load factor gain-scheduled controller. Controllers of reduced-order fixed structures are computed by applying the same H∞ linear design loop-shaping approach for the roll and pitch/yaw-axes. Very good closed-loop performance and robustness properties, which are similar to those provided by full order controllers, are obtained. Finally, the effectiveness of the autopilot augmented by a pure proportional navigation guidance law is verified through a variety of nonlinear trajectory simulations. The latter correspond to nominal flight scenarios with ballistic, non-ballistic stationary, and maneuvering interception points, and to scenarios with perturbed launch conditions or guided projectile dynamics