HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Contribution à l’étude de l’amélioration de la coexistence radio d’un réseau de capteurs sans fils grâce à la mise en œuvre d’algorithme embarqués basés sur l’intelligence artificielle
No full textThe ISM band is an unlicensed band where emission restrictions are relaxed compared with the licensed band. This has led to a proliferation of communication protocols (Wi-Fi, Bluetooth Low Energy (BLE), Zigbee, etc.), despite the absence of any guarantee of spectrum availability. While narrow-band, low-power protocols make it possible to optimise energy for short-range communications and ultimately extend the battery life of IoT devices, their methods of accessing the radio environment have limitations when it comes to guaranteeing good reception conditions in the event of high radio activity in the frequency band used.For this study we focused on improving the interference management of the BLE receiver. Its weaknesses are symptomatic of protocols optimised to be energy-efficient. Wi-Fi uses a relatively high level of power and broadband frequencies, compared to BLE, during transmissions, contributing significantly to BLE interference. We have specifically studied this problem with the case of late arrivals of interference in relation to the freezing of the gain index of the low-noise amplifier (LNA). This configuration is the most dammaging to the optimum compromise between linearity and receiver noise, which is supposed to be guaranteed by the convergence of the LNA gain.The setting up of our study is based on the following observation: due to the random nature of interference, collecting sufficient data on each channel to accurately predict interference on each channel requires very long simulation and measurement times. The type of data set produced in the end will lead to a large model with enormous computational requirements.The first contribution of this study is the evaluation of several machine learning algorithms predicting the LNA gain index or the inability to receive the future packet, using a restricted history of receiver operating metrics and the output is the optimal Automatic Gain Control (AGC) index to be used for the reception of the next packet. As a result, we give the radio the ability to anticipate the appearance of an interferer during payload reception. The final evaluation of the system showed a significant improvement in the Packet Error Rate (PER). In particular, prediction of the AGC makes it possible to broaden the range of packets that can be received.The second contribution is the development of a method for evaluating energy gain as a function of receiver reception statistics. We have provided a method for in-depth analysis of the ML model in order to accurately assess the balance between the energy gain provided by the AGC index prediction and the predicted poor reception/non-detection and the energy consumed by the model to produce the prediction. This method takes into account the energy of AGC index prediction as a function of packet state reception to determine the limit of the ML model of interest in terms of prediction time and average retransmission. The evaluation of the candidate models makes it possible to weight the potential gain and the effective energy consumption of the model.The third contribution concerns the problem of generating a signal with different patterns for training an ML model while avoiding over-fitting, which led us to consider a new cross-validation method. The proposed method makes it easy to define the distribution of the portion of the signal in the training set and the test set, while guaranteeing the randomness of the data allocation and the compatibility between the test set and the training setLa bande ISM est une bande sans licence où les restrictions d'émission sont assouplies par rapport à la bande sous licence. Cela a conduit à une prolifération des protocoles de communication (Wi-Fi, Bluetooth Low Energy (BLE), Zigbee...), malgré l'absence de garantie sur la disponibilité du spectre. Si les protocoles à bande étroite, à faible puissance permettent d'optimiser l'énergie pour les communications à courte portée et à terme d'allonger la durée de vie de la batterie des appareils IoT, leurs méthodes d'accès à l'environnement radio présentent des limites pour garantir de bonnes conditions de réception en cas de forte activité radio dans la bande de fréquence utilisée. Pour cette étude nous nous sommes concentrés sur l'amélioration de la gestion des interférences du récepteur BLE. Ses faiblesses sont symptomatiques des protocoles optimisés pour être économe en énergie. Le Wi-Fi utilise un niveau de puissance relativement élevé et des fréquences à large bande, par rapport au BLE, lors de ces transmissions contribuant de manière significative à l'interférence BLE. Nous avons étudié spécifiquement ce problème avec le cas des arrivées tardive des interférences par rapport au gel de l’index de gain de l’amplificateur faible bruit (LNA). Cette configuration est la plus préjudiciable pour compromis optimal entre la linéarité et le bruit du récepteur qu’est sensé garantir la convergence du gain de l’amplificateur faible bruit (LNA).La mise en place de notre étude repose sur l'observation suivante : en raison du caractère aléatoire des interférences, la collecte de données suffisantes sur chaque canal pour prédire avec précision les interférences sur chaque canal nécessite des temps de simulation et de mesure très longs. Le type d'ensemble de données produit à la fin conduira à un modèle de grande taille avec des exigences de calcul énormes.La première contribution de cette étude consiste en l’évaluation de plusieurs algorithmes de machine learning prédisant l’index de gain du LNA ou l’impossibilité de recevoir le futur paquet, utilisant un historique restreint des métriques de fonctionnement du récepteur et la sortie est l'index de contrôle automatique de gain (AGC) optimal à utiliser pour la réception du paquet suivant. En conséquence, nous donnons à la radio la capacité d'anticiper l'apparition d'un interféreur pendant la réception de la charge utile. L'évaluation finale du système a montré une amélioration significative du Packet Error Rate (PER). Notamment, la prédiction de l'AGC permet d'élargir la gamme des paquets recevables. La deuxième contribution consiste au développement d’un méthode d’évaluation du gain d’énergie en fonction des statistiques de réception du récepteur.Nous avons fourni une méthode d'analyse approfondie du modèle ML afin d'évaluer précisément l'équilibre entre le gain d'énergie fourni par la prédiction de l'indice AGC et la mauvaise réception/non-détection prévue et l'énergie consommée par le modèle pour produire la prédiction. Cette méthode permet de prendre en compte l'énergie de la prédiction de l'indice AGC en fonction de la réception de l'état des paquets pour déterminer la limite du modèle ML d'intérêt en termes de durée de prédiction et de réémission moyenne. L'évaluation des modèles candidats permet de pondérer le gain potentiel et la consommation d'énergie effective du modèle. La troisième contribution concerne la problématique de la génération d'un signal avec différents motifs pour l'entraînement d'un modèle ML tout en évitant le surajustement nous a amenés à envisager une nouvelle méthode de cross validation. La méthode proposée permet de définir facilement la distribution de la portion du signal dans l'ensemble d’entrainement et l'ensemble de test tout en garantissant le caractère aléatoire de l'allocation des données et la compatibilité entre l'ensemble de test et l'ensemble d’entrainemen
Peut-on vraiment co-construire sa recherche ? Enjeux de la recherche autour de population marginalisée à partir de la projection de Visibles
No full textInternational audienc
A digital twin to promote and preserve the endangered volunteer geographic information
No full textInternational audienceAll twins share the same past. But a digital twin does not. Though the Digital Twin Consortium states that Digital Twins (abbr. DT) use “historical data to represent the past and present and simulate predicted futures”, historical heritage data are rarely addressed. It requires access to anything that is preserved digitally: a share comes from institutions today dedicated to this preservation, supplemented, fortunately, by millions of individual contributions, the “volunteer geographic information” (abbr. VGI). We focus on railways. Rail enthusiasts have contributed quantities of original information on features of which few or nothing remains today. Relating features thru time can rely only on geographical locations, and these locations have a past, what requires a time translation. This article proposes a GIS, augmented with tools for accessing and interpreting VGI, giving this GIS the attributes of a digital twin. Rail enthusiasts' VGI is at risk: often published on loosely maintained websites by an aging community of volunteers, the current number of missing links is a testament to what is already lost, at a rate that is accelerating as the contributors age. We aim to add tracking capabilities of endangered VGI to DT, and raise awareness about preserving this VGI in sustainable web archives. The main modules of the DT are continuous enhancing of collected data, quality monitoring, Internet retrieval of heritage-oriented VGI, computer simulations, diachronic comparisons and analyses, and archiving assistance. The architecture is a flexible collection of independent modules, combined as a pyramid whose base is a simple GIS, and communicating thru messages. This architecture allows to add modules easily depending on the intended use of the Digital Twin, including interaction with AI agents
Do this or do that? A model to prioritize reforms
No full textThis paper aims to fill the methodological gap in development economics that until now there exists no quantitative tool that allows to prioritize reforms in a systematic nor optimal way. Following the recent debate on the issues Randomized Control Trials (RCTs) have with establishing external validity and general equilibrium effects, this paper proposes a micro-founded Growth Diagnostics framework to consider general equilibrium effects and prioritize policy prescriptions. Building conceptually on Hausmann et al. (2005), we set up two continuous-time Overlapping Generations (OLG) models to obtain the private and social net-marginal valuations of economic activities as measured by their shadow values. With these in hand, we define the wedges in the net-marginal private and social valuations to set up a new planner problem (the super policy maker problem), where the planner minimizes an aggregate function of disagreement as measured by the wedges. We illustrate our frame-work with an application to the literature on structural change. Worth noting, the final wrapping optimization problem allows to prioritize optimally economic reforms in a second-best framework, thus, to put it in the words of Rodrik (2010), to first diagnose before one prescribes the remedy
Organics in the stratosphere: new insights from weather balloon flights conducted in France
No full textInternational audienceAtmospheric aerosols significantly influence the Earth’s climate, yet they remain a major source of uncertainty in climate projections. Here, we present a set of unique offline analysis of organics collected from the ground to the stratosphere with ultimate sensitivity using a new balloon-borne sampling technique deployed on simple weather balloons. Activated and functionalized carbon fabrics were used to target chemical compounds in the gaseous and aerosol forms during three flights over France. Ultra-high-resolution mass spectrometry, enabling the detection of thousands of organic compounds, and highlighted high-carbon-number organics, indicating oligomerization in long-lived tropical stratospheric air masses. Our measurements suggest that large stratospheric sulfate aerosols from volcanic eruptions could play a role in the formation of secondary aerosols in internally mixed particles and may explain the presence of organics with higher carbon numbers. These measurements revealed a drop in the number of unsaturated carbons at higher altitudes
The new nature of stations
No full textInternational audienceSymbols and legacies of a prestigious past, railway stations and their neighbourhoods are now the focus of strategies for adapting to global change. The book explores the various technical processes involved in this approach (reuse of building materials for renovation, environmental and landscape integration, energy transformations, etc.) and shows how these processes link railway regeneration to territorial development. The twenty-one experts assembled in this book deploy critical approaches and international perspectives to think of railway stations as transdisciplinary border objects where naturalistic, architectural and political perspectives can be articulated. As such, the book will enlighten users, as well as residents and citizens, on the processes of socio-technical transitions underway, as much as it will help them question the costs, effects and ends of the eco-modernization of our public goods and services
Weak solutions of stochastic volterra equations in convex domains with general kernels
No full textWe establish new weak existence results for d-dimensional Stochastic Volterra Equations (SVEs) with continuous coefficients and possibly singular one-dimensional nonconvolution kernels. These results are obtained by introducing an approximation scheme and showing its convergence. A particular emphasis is made on the stochastic invariance of the solution in a closed convex set. To do so, we extend the notion of kernels that preserve nonnegativity introduced in Alfonsi (2025) to non-convolution kernels and show that, under suitable stochastic invariance property of a closed convex set by the corresponding Stochastic Differential Equation, there exists a weak solution of the SVE that stays in this convex set. We present a family of non-convolution kernels that satisfy our assumptions, including a nonconvolution extension of the well-known fractional kernel. We apply our results to SVEs with square-root diffusion coefficients and non-convolution kernels, for which we prove the weak existence and uniqueness of a solution that stays within the nonnegative orthant. We derive a representation of the Laplace transform in terms of a non-convolution Riccati equation, for which we establish an existence result
Plastic debris dataset on the Seine riverbanks: up to 38 000 pre-production plastic pellets reported per square meter
No full textInternational audiencePlastic pollution in rivers is a major source for plastic pollution into the ocean. However, it is now recognized that plastics may accumulate in rivers for years, especially in estuaries, before reaching the ocean. This long residence time favours fragmentation of macroplastics into smaller and smaller pieces, but relative data are still carse. Here we present data from the downstream part of the Seine estuary in a historical deposit full of plastic debris, with the highest concentration of industrial plastic pellet ever reported in France. Plastic debris (down to 6 mm according to sieving limits) were classified using the updated European classification J-list. The sampled site is located close to the river mouth within a national natural reserve, surrounded by international harbour activities and two major industrial plastic producers: Total Energy and Exxon Mobil. A surface of only 1 m2 was sampled in a visual maximum of plastic pollution. Over 100 000 plastic debris were counted or estimated when it comes to plastic debris <6 mm. Items were classified and weighted by category for a total mass higher than 4 kg. By mass (count), 24% (38%) of total plastic debris were pre-production plastic pellets, 21% (1%) were unidentified, colourful, plastic fragments (≥2.5 cm), and 19% (33%) were unidentified, colourful, plastic fragments (<2.5 cm)
Social cues for experimenter incompetence influence choice blindness
No full textInternational audienceChoice blindness refers to a surprising blind spot we have about choices made only seconds ago. After making a choice between two items, observers presented with the unchosen item may fail to report the incongruence, and even provide justifications for a choice they did not make. Here, we show that this effect is modulated by participant’s perception of the reliability of the environment. In three experiments, we introduced cues about the competence or incompetence of experimenters, either during or before the traditional choice blindness phase. When manifest reliability of the experimenter decreased, participants were more likely to report the mismatch between the chosen item and the item presented to them. Our results reinforce the notion that choice blindness is a context-dependent phenomenon, permeable to social cues in the context of psychological experiments. Dataset and the analysis scripts are available at the Open Science Foundation at: https://osf.io/ht769/
Low emission zone and inequalities: car restriction and reduction of air pollution exposure in the Paris region
No full textInternational audienceOver the past few decades, Low Emission Zones (LEZs) have emerged as a widely adopted tool for improving air quality accross Europe. However, the scientific literature devotes limited attention to the inequalities associated with their implementation. In addressing this gap, this article proposes to identify the residents least affected by reductions in air pollution and those most impacted due to ownership of non-compliant vehicles, focusing on the Paris region (France). By integrating data from the Enquête Globale Transport with data on spatio-temporal NO2 concentration levels, the methodological approach first evaluates the spatio-temporal exposure of surveyed individuals, considering their activities and transport modes. Thus, the objective is to study individual inequalities related to reduced exposure levels and non-compliant vehicle ownership. A series of logistic regressions model independently, and then subsequently jointly, estimate the likelihood of individual exposure levels falling below the WHO recommended threshold and the need to replace a non-compliant vehicle. The results reveal that the poorest residents are the most affected by restrictions on non-compliant vehicles, both in terms of exposure and ownership, and ultimately facing a double inequality. This inequality is less evident in the case of a more ambitious scenario