Ecole Nationale de l'Aviation Civile
Toutes les publications de l'ENAC en direct.
All ENAC publications.
This paper presents a robust cascaded control architecture for fully actuated and overactuated multirotors. The proposed approach combines Incremental Nonlinear Dynamic Inversion and structured ℋ∞ control, and extends their application beyond underactuated platforms considered in previous work. It reduces dependence on an accurate multirotor model by shifting the modeling effort toward the actuator dynamics, thereby improving closed-loop robustness in the presence of uncertainties. Moreover, attitude and position tracking are achieved through a geometric weighted least-squares control allocation scheme formulated as a quadratic optimization problem in the guidance loop...
Une caractérisation expérimentale de l’écoulement généré par une aile battante déformée activement est proposée. La déformation active est réalisée par des actionneurs piézoélectriques installés à l’intérieur de l’aile permettant un contrôle précis de la flexion de cette dernière. En déphasant le signal électrique par rapport au mouvement de battement de l’aile, la déformation en flexion de l’aile le devient aussi. Ceci permet alors de caractériser l’influence du déphasage de la flexion δθ sur l’écoulement et donc les performances aérodynamiques obtenues pour une aile battante. Ce travail se concentre sur deux valeurs du déphasage : δθ = π/2 et δθ = 3π/2. Dans les deux...
In granular materials, cohesive forces, notably from van der Waals interactions, are expected to play an important role in low gravity. Theoretical estimates predict that on small bodies, van der Waals forces may become larger than the weight of large boulders [1]; the presence of fines might further increase the van der Waals forces, thus increasing the global cohesion of asteroid material [2]. While the exact magnitude of the cohesive forces is under debate [3], scaling of the different forces in presence clearly indicates that cohesion should be considered when studying small bodies and designing simulants for laboratory testing The latter is a particular challenge, as...
Designing space systems involves navigating vast trade spaces while balancing competing cost, perfor- mance, and feasibility constraints. Traditional aerospace design often follows a point-based approach, in which a single concept is refined iteratively. However, this risks premature convergence on suboptimal solutions due to early-stage uncertainty and system complexity. Set-Based Design (SBD) offers a paradigm shift by enabling the parallel exploration of multiple feasible solutions before gradually converging toward an optimal design. This study applies SBD principles to the development of low-cost lunar landers, inte- grating Multi-Attribute Tradespace Exploration...
Nowadays, as space exploration advances, forthcoming Moon missions have gathered significant attention from both the public and scientists. Due to this increased interest and effort in returning to the Moon, there is a growing need to comprehend the difficulties faced by human operators during teleoperation tasks in extreme environments. Understanding teleoperation performance is crucial for future missions, as it enables the development of reliable and efficient communication systems between human operators and robotic systems. This is essential for executing complex tasks, conducting experiments, and addressing unforeseen challenges in the remote and harsh environments...
A data-driven surrogate model is developed to emulate neighboring coaxial injectors. Based on U-Nets, the model leverages transfer learning to limit the computational cost of generating a large database of multi-injector simulations. A design of experiment is conducted to generate two datasets of two-dimensional Reynolds-Averaged Navier-Stokes (RANS) simulations, sampling recess length and flame confinement length, used as geometric inputs to the networks. The model is first pre-trained on simulations of single injectors and later fine-tuned on simulations of two interacting injectors. The results demonstrate the surrogate model’s ability to replicate complex interactions...
New sensing and display hardware makes it increasingly easy to incorporate live data-driven feedback and visualizations into sports and physical activities. For example, a jogger using a smartwatch or smart glasses can monitor their pace and heart rate in-situ and adjust their behavior to meet their goals. However, the diverse contexts in which physical activities take place make designing for such experiences very different from designing traditional desktop-based visualizations. Activities may occur in dynamic environments (outdoors in the rain versus indoors at a gym), involve different social dynamics (volleyball versus solo hiking), and afford different input and...
BSS face a persistent operational challenge, spatial and temporal imbalances between stations that force operators into costly manual redistribution. Existing demand-forecasting models, from count data regressions to deep learning architectures, treat each time slot independently and overlook a structural feature of BSS usage the dependency between morning and evening flows. This paper formalizes that dependency through a Hurdle Negative Binomial-Binomial model applied to the VélôToulouse network. For each origin station, the model estimates p, the probability that a morning trip generates a same day evening return, using aggregated flows without individual level data...
Decoding electroencephalography (EEG) signals with deep learning remains challenging due to strong interand intra-individual variability, low signal-to-noise ratio, and the absence of an optimal architecture. While deep neural networks have demonstrated promising performance for EEG decoding, selecting an appropriate model architecture remains a costly and subject-specific process. Neural architecture search approaches partially address this issue but are computationally prohibitive in practice. In this work, we investigate a growing neural network strategy that adapts model expressivity during training, only expanding the architecture when required by the data. Building...
