HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Review of Autonomous Racing Competitions and Technologies
No full textInternational audienceThis work comprehensively analyzes autonomous racing competitions and their technologies, addressing a gap in current literature. We evaluate these platforms through five key metrics: cost, system modularity, speed performance, technical complexity, and academic impact. Using bibliometric analysis of scientific literature, we identify dominant technologies, including Model Predictive Control and reinforcement learning. Our comparative framework reveals clear trade-offs: Formula Student Driverless offers accessibility and academic involvement but features lower speeds. At the same time, full-scale competitions like Indy Autonomous Challenge provide high-velocity testing but at increased cost. Reduced-scale platforms serve as valuable precursors to full-scale projects, offering cost-effective insights into perception, control, and system integration. The study demonstrates how competitions complement each other in advancing autonomous driving technologies and identifies patterns in simulator usage for algorithm development and validation. This comparison provides researchers with guidance for selecting appropriate platforms based on research objectives and available resources
SpeCT: A state-of-the-art tool to calculate correlated-k tables and continua of CO2-H2O-N2 gas mixtures
No full textInternational audienceA key challenge in modeling (exo)planetary atmospheres lies in generating extensive opacity datasets that cover the wide variety of possible atmospheric composition, pressure, and temperature conditions. This critical step requires specific knowledge and can be considerably time-consuming. To circumvent this issue, most available codes approximate the total opacity by summing the contributions of individual molecular species during the radiative transfer calculation. This approach neglects inter-species interactions, which can be an issue for precisely estimating the climate of planets. To produce accurate opacity data, such as correlated-k tables, χ factor corrections of the far wings of the line profile are required. We propose an update of the χ factors of CO2 absorption lines that are relevant for terrestrial planets (pure CO2, CO2-N2, and CO2-H2O). These new factors are already implemented in an original user-friendly open-source tool, named SpeCT, designed to calculate high-resolution spectra. This tool produces correlated-k tables for mixtures made of H2O, CO2, and N2, and accounts for inter-species broadening. In order to facilitate future updates of these χ factors, we also provide a review of all the relevant laboratory measurements available in the literature for the considered mixtures. Finally, we provide in this work eight different correlated-k tables and continua for pure CO2, CO2-N2, CO2-H2O, and CO2-H2O-N2 mixtures based on the MT_CKD formalism (for H2O), and calculated using SpeCT. These opacity data can be used to study various planets and atmospheric conditions, such as Earth’s paleo-climates, Mars, Venus, Magma ocean exoplanets, and telluric exoplanets
On polarization, incommensurability, and value-laden research. A response to Bjørn Hofmann, 2024
No full textInternational audienceIn this commentary, I integrate Bjørn Hofmann's thorough analysis of polarization in research with two considerations. First, Hofmann defines polarization as characterized by incommensurable positions. This makes his definition too strict, as hardly any disagreement in modern science, including the cases he discusses, is based on genuine incommensurability. Polarization in research is better characterized in terms of perceived incommensurability between opposite groups. This is not a mere terminological issue. In the absence of genuine incommensurability, talking about incommensurability to describe polarized debates only risks exacerbating them. Second, Hofmann reviews several explanations of polarization but includes only value differences in his definition. Because values are ubiquitous in research, the role of values in polarization should be better qualified. Hofmann's current definition risks suggesting that values are a special feature of polarization, rather than a common feature of scientific research.Switching from the incommensurability to the perceived incommensurability criterion would make Hofmman's definition more precise. Better qualifying the role of values in polarization would make it more consistent with the values in science literature and his own analysis. Both tweaks will help forestall possible risks in communication that could hinder attempts to smooth over polarized debates, including those attempts reviewed by Hofmann
Apport des nouvelles techniques analytiques pour le suivi des substances indésirables dans les urines et les sols
No full textNational audienc
Biogeographical Regions and Climate Change: Lanternfishes Shed Light on the Role of Climatic Barriers in the Southern Ocean
No full textInternational audienceTo predict the spatial responses of biodiversity to climate change, studies typically rely on species-specific approaches, such as species distribution models. In this study, we propose an alternative methodology that investigates the collective response of species groups by modelling biogeographical regions. Biogeographical regions are areas defined by homogeneous species compositions and separated by barriers to dispersal. When climate acts as such a barrier, species within the same region are expected to respond similar to changing climatic conditions, enabling the prediction of entire region shifts in response to future climate scenarios. We applied this approach to the Southern Ocean, which exhibits sharp climatic transitions known as oceanic fronts, focusing on the mesopelagic lanternfishes (family Myctophidae). We compiled occurrence data for 115 lanternfish species from 1950 onwards and employed a network-based analysis to identify two major biogeographical regions: a southern and a subtropical region. These regions were found to be distinct, with minimal overlap in species distributions along the temperature gradient and a separation around 8°C, indicating that temperature likely acts as a climatic barrier. Using an ensemble modelling approach, we projected the response of these regions to future temperature changes under various climate scenarios. Our results suggest a circumpolar expansion of the subtropical region and a contraction of the southern region, with the Southern Ocean becoming a cul-de-sac for southern species. Ultimately, our results suggest that when support is found for the climatic barrier hypothesis, community-level models from a 'group first, then predict' strategy may effectively predict future shifts in species assemblages.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.RÉSUMÉ Les études prédictives sur les réponses spatiales de la biodiversité au changement climatique reposent généralement sur des approches espèce par espèce, comme les modèles de distribution d’espèces. Dans cette étude, nous proposons une méthodologie alternative qui explore la réponse collective de groupes d’espèces en modélisant des régions biogéographiques. Les régions biogéographiques sont des zones définies par une composition spécifique homogène et séparées par des barrières à la dispersion. Lorsque le climat agit comme une barrière, on s’attend à ce que les espèces d’une même région répondent de manière similaire à des changements de conditions climatiques, permettant ainsi de prédire les déplacements de régions entières en réponse à des scénarios climatiques futurs. Nous avons appliqué cette approche à l’Océan Austral, qui présente d’abruptes transitions climatiques correspondant à des fronts océaniques, en nous concentrant sur les poissons‐lanternes mésopélagiques (famille Myctophidae). Nous avons compilé des données d’occurrence pour 115 espèces de poissons‐lanternes depuis 1950, et effectué une analyse basée sur les réseaux pour identifier deux régions biogéographiques majeures : une région australe et une région subtropicale. Celles‐ci se sont révélées distinctes, avec un chevauchement limité entre les distributions d’espèces le long du gradient de température et une séparation autour de 8°C, ce qui indique que la température pourrait agir comme une barrière climatique. En appliquant une approche de modélisation d’ensemble, nous avons projeté la réponse de ces régions en fonction de futurs changements de température selon divers scénarios climatiques. Nos résultats suggèrent une expansion circumpolaire de la région subtropicale et une contraction de la région australe, où l’Océan Austral formerait un cul‐de‐sac pour les espèces polaires. Enfin, nos résultats suggèrent que, lorsque l’hypothèse d’une barrière climatique est vérifiée, les modèles de communautés suivant la stratégie “assembler, puis prédire” pourraient effectivement prédire les potentiels déplacements d’assemblages d’espèces
Enhancing Soil Moisture Statistical Retrieval from SMOS using Partial Convolutions and Localization Strategies
No full textInternational audienceNeural networks have been used for the retrieval of soil moisture (SM) from microwave observations over the last 20 years. The exploitation of the SMOS (Soil Moisture and Ocean Salinity) observations has largely benefited from such statistical models. However, these retrievals are currently done at the pixel level, ignoring spatial context and using a constant incidence angle configuration approach. While pixels are seen with a varying number of angles from the SMOS instrument, only pixels monitored by a fixed pre-selected angle configuration are considered. These two limitations can have a negative impact on the quality of the retrievals. This paper introduces a new NN architecture that combines two powerful innovations. Firstly, the new model is image-based: it ingests entire SMOS orbit swath and thus leverages the strong spatial pattern present in the satellite observations. Secondly, a “partial convolutional layer” is tested. It allows being flexible, in the retrieval, on the angle configuration: more incidence angles can be exploited when they are available. Finally, a concept called “Localization” is also exploited, helping the NN retrieval to adapt his behaviour to specific local conditions Experiments are conducted at SMOS orbit scale over the contiguous united states (CONUS) region using five years of SMOS data (2016–2019). A temporal correlation of 0.74 (unit-less) with respect to ERA5 reanalysis and 0.62 with respect to in situ SM measurements network are obtained (to be compared to respectively 0.63 and 0.60 with the legacy pixel and fixed angle based approach). Furthermore, the use of partial convolutions results in enlarging the retrieval domain by +240% versus legacy retrieval, and by +140% versus the operational SMOS Level-2 product
Nonlinear weak error expansion of McKean-Vlasov stochastic differential equations
No full textAccording to Talay and Tubaro \cite{talay_expansion_1990}, the weak error between the solution to a stochastic differential equation with smooth coefficients and its Euler-Maruyama scheme can be expanded in powers of the time-step. In the present paper, we generalize this result to the case when the error is measured by a smooth functional on the Wasserstein space of probability measures in place of the linear functional given by the expectation of a smooth function considered in \cite{talay_expansion_1990}. Since this does not complicate our analysis based on the master partial differential equation, we even deal with the McKean-Vlasov case when the coefficients of the stochastic differential equation may depend on its current marginal distribution
Large-scale experimental investigation of the reliability of confidence measures
No full textInternational audienceWhether individuals feel confident about their own actions, choices, or statements being correct, and how these confidence levels differ between individuals are two key primitives for countless behavioral theories and phenomena. In cognitive tasks, individual confidence is typically measured as the average of reports about choice accuracy, but how reliable is the resulting characterization of within- and between-individual confidence remains surprisingly undocumented. Here, we perform a large-scale resampling exercise in the Confidence Database (103 studies, 6000 participants) to investigate the reliability of individual confidence estimates, and of comparisons across individuals’ confidence levels. Our results show that confidence estimates are more stable than their choice-accuracy counterpart, reaching a reliability plateau after roughly 50 trials, regardless of a number of task design characteristics. While constituting a reliability upper-bound for task-based confidence measures, and thereby leaving open the question of the reliability of the construct itself, these results characterize the robustness of past and future task designs
Assessment of groundwater impact on the water temperature of small sand-pit lakes through one-dimensional modelling
No full textInternational audienceIn urban areas, small lakes provide many ecosystem services including biodiversity, landscape composition, cooling islands, recreation, etc. Many have been created during the recent decades as sand-pit lakes. Their origin comes from urbanisation which requires large quantities of sand and gravel for the construction of buildings and infrastructure. Gravel is often extracted from riverbeds, beach deposits or alluvial fans. When gravel extraction stops, the quarries fill up with groundwater and become artificial lakes.The thermal regime and hydrodynamics of these lakes have a strong influence on their ecological functioning and on the fate of contaminants in the water column. In order to better understand their physical behaviour and to which extent it may be affected by climate change, numerical modelling can be very effective.Calibration of the model parameters is a crucial step to obtain reliable modelling results. However, the available field data are generally too scarce to obtain a single set of parameter values. Performing a sensitivity analysis allows to identify the most sensitive parameters that need to be calibrated.The results of the parameter sensitivity analysis and the calibration of a one-dimensional model (GLM, General Lake Model) (Hipsey et al., 2019) are presented. The sensitivity analysis was performed according to a global sensitivity analysis technique, the Morris method (Herman et al., 2013). For the parameter calibration, the CMA-ES method (Covariance Matrix Adaptation - Evolution Strategy), which has been previously used for GLM lake modelling (Ladwig et al., 2021), was applied. The study site is a sand-pit lake located in the Great Paris region. High-frequency water temperature records are available at 4 depths in the water column for the last two years (2023 and 2024).The sensitivity analysis showed that the thermal regime of the lake is particularly sensitive to the values of 4 parameters that are related to the meteorological forcing (sw, a scaling factor to adjust the shortwave radiation data), the light attenuation in the water column (Kw), the latent heat flux transfer coefficient (Ce) and the mean sediment temperature (sed_temp_mean). Calibration of these 4 parameters was then conducted. The simulation obtained with the calibrated parameters was then compared with a reference simulation performed using default values for all parameters.The results highlight the importance of including the sediment temperature to correctly simulate the temperature of the lake bottom layers. The high-frequency monitoring (time step = 15mn) allows to accurately check the efficiency of the calibration method. After the calibration of the 4 parameters identified in the sensitivity analysis, the simulation of the lake temperature is significantly improved, according to different metrics (e.g. RMSE decreasing from 1.8°C to 0.8°C)
Improved L^2-error estimates for the wave equation discretized using hybrid nonconforming methods on simplicial meshes
No full textWe present improved L^2-error estimates on the time-integrated primal variable for the wave equation in its first-order formulation. The space discretization relies on a hybrid nonconforming method, such as the hybridizable discontinuous Galerkin, the hybrid high-order or the weak Galerkin methods. We consider both equal-order and mixed-order settings on simplices, and include the lowest-order case with piecewise constant unknowns on the faces and in the cells. Our main result is a superclose, resp., optimal bound on the above error in the equal-, resp., mixed-order case. A key result of independent interest to achieve these estimates are novel approximation estimates for an interpolation operator inspired from the hybridizable discontinuous Galerkin literature