HAL-Polytechnique
Not a member yet
51406 research outputs found
Sort by
Experimental validation of scenario-based stochastic model predictive control of nanogrids
No full textInternational audienceIn microgrids and nanogrids, challenges arise from the inherent intermittency of renewable energy sources and the need to meet uncertain energy demand from users. To address these uncertainties, this paper investigates a two-layer, scenario-based stochastic Model Predictive Control (MPC) for a real lab-scale photovoltaic (PV)-based nanogrid. The high-level layer, which operates slowly and over longer time horizons, computes optimal reference values for the low-level layer based on predictions of uncertainty in PV generation and consumer load. The low-level layer, which operates on shorter time horizons and at higher frequencies, relies on scenario-based MPC. Scenario-based MPC has several advantages, such as not requiring prior knowledge of the underlying probability distribution. However, it can suffer from significant computational burdens, especially in real-time applications like nanogrid control. To overcome this challenge, this paper employs the Alternating Direction Method of Multipliers (ADMM) to efficiently solve the optimization problem. First, real PV and load data are used to characterize the scenarios. Then, the proposed scheme is experimentally validated on a PV-based nanogrid. The results show that the two-layer scenario-based MPC outperforms the two-layer chance-constrained MPC and significantly improves performance compared to a rule-based energy management system
A Continuation Method Based on CMA-ES
No full textInternational audienceIn this poster, we showcase a new algorithm for approximating the Pareto set of two-objective (unconstrained) optimization problems based on the idea of continuation. The algorithm tries to move "along" the Pareto set from one single-objective optimum to the other and back via a single-objective reformulation of the two-objective problem and the well-known CMA-ES as single-objective solver. The introduced algorithm BOG-CMA-ES (standing for bi-objective gradient based CMA-ES) is visually analyzed on simple convex-quadratic objective functions and extensively benchmarked on the bbob-biobj test suite of the COCO platform, including comparisons with current state-of-the-art algorithms
Open Problem: Two Riddles in Heavy-Ball Dynamics
No full textThis short paper presents two open problems on the widely used Polyak's Heavy-Ball algorithm. The first problem is the method's ability to exactly \textit{accelerate} in dimension one exactly. The second question regards the behavior of the method for parameters for which it seems that neither a Lyapunov nor a cycle exists. For both problems, we provide a detailed description of the problem and elements of an answer
The Vanilla Sequent Calculus is Call-by-Value
No full textInternational audienceExisting Curry-Howard interpretations of call-by-value evaluation for the λ -calculus are either based on ad-hoc modifications of intuitionistic proof systems or involve additional logical concepts such as classical logic or linear logic, despite the fact that call-by-value was introduced in an intuitionistic setting without linear features. This paper shows that the most basic sequent calculus for minimal intuitionistic logic—dubbed here vanilla —can naturally be seen as a logical interpretation of call-by-value evaluation. This is obtained by establishing mutual simulations with a well-known formalism for call-by-value evaluation
Iodine plasmas for space propulsion and industrial applications
No full textInternational audienceWith as many as 2000 satellites per year forecast to be launched over the next decade, onboard propulsion systems will become increasingly important for ensuring both mission success and a sustainable space environment. Plasma-based electric propulsion systems are particularly attractive because of their high fuel efficiency, but due to challenges with conventional propellants such as xenon, a strong interest in viable alternatives has emerged. One such alternative is iodine, which in addition to space-based applications, is also of use in a number of ground-based industrial applications such as plasma etching. With a lower cost, higher global production output, and a reduced ionization threshold compared with xenon, iodine has the potential to meet current and future space industry demand while also providing improved propulsion performance. Furthermore, iodine is a solid at typical ambient conditions with a high storage density. However, iodine is chemically reactive with many common materials and has a more complex plasma chemistry that includes molecular dissociation, attachment to form negative ions, and several ionization processes creating positive atomic and molecular ions. This topical review provides a comprehensive overview of iodine within the context of plasma applications and also serves as a useful data source for various thermodynamic properties, collision cross-sections, and iodine-surface interactions. In addition to discussing the physical and atomic/molecular properties of iodine, we also highlight important theoretical, numerical, and experimental work in the field and discuss the current state-of-the-art: including the space flight heritage of iodine-fueled propulsion systems and remaining research/technical challenges
FourieRF: Few-Shot NeRFs via Progressive Fourier Frequency Control
No full textInternational audienceWe present a novel approach for few-shot NeRF estimation, aimed at avoiding local artifacts and capable of efficiently reconstructing real scenes. In contrast to previous methods that rely on pre-trained modules or various data-driven priors that only work well in specific scenarios, our method is fully generic and is based on controlling the frequency of the learned signal in the Fourier domain. We observe that in NeRF learning methods, highfrequency artifacts often show up early in the optimization process, and the network struggles to correct them due to the lack of dense supervision in few-shot cases. To counter this, we introduce an explicit curriculum training procedure, which progressively adds higher frequencies throughout optimization, thus favoring global, low-frequency signals initially, and only adding details later. We represent the radiance fields using a grid-based model and introduce an efficient approach to control the frequency band of the learned signal in the Fourier domain. Therefore our method achieves faster reconstruction and better rendering quality than purely MLP-based methods. We show that our approach is general and is capable of producing high-quality results on real scenes, at a fraction of the cost of competing methods. Our method opens the door to efficient and accurate scene acquisition in the few-shot NeRF setting
The Geography of Racialized Commerce and Gentrification
No full textInternational audienceThis paper demonstrates the significance of considering organizational heterogeneity and racial dynamics, pushing beyond the market-based approaches that primarily consider socioeconomic differences to explain disparities in organizational resources across neighborhoods. Drawing on the business directory data harmonized with the Census and American Community Survey data in 10 different US cities spanning from 2000 to 2010, my findings show that while essential businesses are relatively evenly distributed across neighborhoods with varying ethnoracial compositions, discretionary businesses are disproportionately concentrated in white neighborhoods, net of their socioeconomic status. By contrast, discretionary businesses are consistently underrepresented in black neighborhoods, even when they are socioeconomically well-off. Furthermore, socioeconomic upgrading of the neighborhood is associated with an increase in discretionary businesses, particularly when gentrification accompanies white influx. Gentrification without white influx did not bring about significant growth in discretionary businesses, suggesting that discretionary businesses not only symbolize middle-class markers but also are racialized as white
When Lions meets Krugman: A mean-field game theory of spatial dynamics
No full textWe propose a mean-field game (MFG) set-up to study the dynamics of spatial agglomeration in a continuous space-time framework where trade across locations may follow a broad class of static gravity models. Forward-looking intertemporal utility-maximizing agents work and migrate in a twodimensional geography and face idiosyncratic shocks. Equilibrium wages and prices depend on their common distribution and adjust statically according to the underlying trade model. We first prove existence and uniqueness of the static trade equilibrium. We then prove existence of dynamic equilibria. In the case of Krugman (1996)'s racetrack economy, we obtain closed-form solutions for small sinusoidal perturbations around the steady state, and we identify the sets of parameters that lead to agglomeration or dispersion. We exploit the MFG structure of the model to explicitly quantify how uncertainty and forward-looking expectations contribute to agglomeration and dispersion. In particular, we show that, regardless of the static trade model, forward-looking expectations always promote agglomeration, but cannot reverse the dominant pattern that would arise under myopic behavior
Coherent Tietze Transformations of 1-Polygraphs in Homotopy Type Theory
No full textInternational audiencePolygraphs play a fundamental role in algebra, geometry, and computer science, by generalizing group presentations to higher-dimensional structures and encoding coherence for those. They have recently been adapted by Kraus and von Raumer to the setting of homotopy type theory, where they are useful to define and study higher inductive types. Here, we develop the theory of 1-dimensional polygraphs, which correspond to presentations of sets in homotopy type theory. This requires us to introduce a dedicated notion of Tietze transformation, generalizing their well-known counterpart in group theory: the equivalence generated by those transformations characterizes situations where two 1-polygraphs present the same set. We also show a homotopy transfer theorem, which provides a way to transport coherence structures from one 1-polygraph to another. This work lays the foundations for a general theory of polygraphs in arbitrary dimensions, which should be useful for instance to define and study coherent group presentations, allowing for synthetic (co)homology computations. Most of the results in the article have been formalized with the Agda proof assistant using the cubical HoTT library
Planification de trajectoires en présence d’obstacles, prédiction et ordonnancement temps-réel
No full textDespite the growing interest, fully driverless operation of autonomous vehicles (AVs) remains limited. In this context, our research developsmethods that enable AVs to operate safely while efficiently managing their hardware. We focus on twocore sources of uncertainty: (i) the behavior of surrounding vehicles and (ii) the variable execution timesof tasks within the autonomous system. The former relates to prediction, which is essential for safety, while the latter concerns real-time system operation, crucial for efficient resource allocation and timelyreactions. To address the prediction problem, we leverage recentadvances in conformal prediction, a theory for uncertainty quantification that provides prediction zoneswith formal probabilistic guarantees. In this context,we introduce ConForME, a method for computing tight prediction zones around multi-step predictions from neural networks. ConForME advances the state of theart by reducing prediction zone sizes by up to 52%. Building on this, we propose PROSPERS, a framework that combines ConForME with mixed-criticality scheduling. By switching between criticalitymodes based on probabilistic trajectory predictions,PROSPERS ensures safe mode transitions. This approach not only improves resource allocation, but alsobridges a key gap by addressing both uncertaintiessimultaneously. Additionally, we develop a Rapidly-Exploring Random Tree (RRT)-based planning algorithm that computes dynamically feasible short-termplans and provide a proof of its probabilistic completeness.Alongside PROSPERS, we present a modular ROS 2implementation, which is compatible with Common-Road scenarios and we use it to simulate safe overtaking maneuvers. In conclusion, this thesis providesa comprehensive perspective on probabilistic safetyand efficient resource allocation for AVs, paving theway for more reliable and adaptable autonomous driving systems in complex real-world environments.Malgré l’intérêt croissant pour les véhicules autonomes (AVs), leur fonctionnement entièrement sans conducteur reste limité. Dans ce contexte, notre recherche développe des méthodes qui permettent aux AVs de fonctionner en sécurité tout en gérant efficacement leurs ressources matérielles.Nous nous concentrons sur deux sources principales d’incertitude : (i) le comportement des véhicules environnants et (ii) les temps d’exécution variables des taches au sein du systeme autonome. La première est liée à la prédiction, essentielle pour la sécurité, tandis que la seconde concerne le fonctionnement des systèmes temps réel, crucial pour une allocation efficace des ressources et des réactions rapides.Pour aborder le problème de la prédiction, nous exploitons les avancées récentes en matière de prédiction conforme, une théorie de quantification de l’incertitude qui fournit des zones de prédiction avec des garanties probabilistes formelles. Dans ce contexte, nous introduisons ConForME, une méthode pour calculer des zones de prédiction serrées autour des prévisions multi-horizon de réseaux de neurones.ConForME fait progresser l’état de l’art en réduisant la taille des zones de prédiction jusqu’à 52 %. En nous appuyant sur ce résultat, nous proposons PROSPERS, un cadre de travail qui combine ConForME avec l’ordonnancement à criticité mixte.En alternant entre les modes de criticité sur la base de prédictions probabilistes de trajectoire, PROSPERS assure des transitions de mode sûres. Cette approche non seulement améliore l’allocation des ressources, mais comble également une lacune majeure en traitant simultanément les deux incertitudes. De plus, nous développons un algorithme de planification basé sur les Arbres à Exploration Rapide Aléatoires(RRTs) qui calcule des plans à court terme dynamiquement faisables et nous fournissons une preuve de sa complétude probabiliste.Parallèlement à PROSPERS, nous présentons un implémentation modulaire sous ROS 2, qui est compatible avec les scenarios CommonRoad et que nous utilisons pour simuler des manœuvres de dépassement sûres. En conclusion, cette thèse offre une perspective complète sur la sécurité probabiliste et l’allocation efficace des ressources pour les AVs,ouvrant la voie à des systèmes de conduite autonome plus fiables et adaptables dans des environnements réels complexes