Search

TensorFlow implementation of R2D2 from "Meta-learning with differentiable closed-form solvers" (ICLR 2019)

In this paper, we explore the security of a distributed model predictive control (DMPC) scheme. A global system is decomposed into several local subsystems which share a common resource. For instance a heating problem among several rooms with a limited amount of power. Each subsystem owns a model predictive control (MPC) controller which cooperates and works iteratively with all the other controllers in order to reach the system's objective. Here, we focus on the security of such a system. Indeed, we report that a user could disturb the DMPC through several attacks in order to destabilize the resource distribution. Hence, examples of cheating protocols and counter-cheat ideas are presented and discussed.

ROBOTICS; In this small paper, we explore the different kind ofMesh network protocols as of today. The objective is to give tothe reader an overview of the different existing techniques andto decide which one would be the most efficient for a roboticapplication. Indeed the final goal is to realize a Mesh networkthat would enable several robots to communicate. This paperis a state of the art of Mesh application for robotics as well asa experiment review of the use of mesh network for robotics.

National audience; Contexte : l’automatique est la branche des sciences de l’ingénieur qui s’intéresse au pilotage automatisé des systèmes, par exemple la régulation de vitesse d’une voiture ou de la température d’une pièce. En automatique, les développements d’algorithmes se font très souvent dans l’environnement commercial Matlab/Simulink, tant en enseignement qu’en recherche. Cependant, beaucoup des briques de base pour travailler sur des algorithmes d’automatique existent en Python.Cet exposé est une introduction à une méthode très puissante et très usitée, la commande prédictive, en Python. Également connue sous acronyme anglais MPC (Model Predictive Control), cette méthode permet de commander des systèmes complexes à l’aide de méthodes d’optimisation.Nous utiliserons l’exemple du pilotage optimal du chauffage d’un bâtiment de façon à minimiser la consommation d’énergie. Nous décrirons les principales étapes de la mise en œuvre de la commande en Python (notebook Jupyter mis en ligne https://github.com/pierre-haessig/mpc-pyconfr-2016). Nous présenterons les principales bibliothèques utilisées (e.g. numpy), et plus particulièrement les routines d’optimisation de cxvopt (http://cvxopt.org/).Perspectives : un des intérêts d’utiliser Python et des bibliothèques libres est de permettre d’embarquer facilement l’algorithme créé sur une plateforme du type Raspberry Pi. Ainsi, l’exemple présenté du pilotage d’un chauffage pourrait être intégré dans des plateformes de domotique ouvertes.

International audience

International audience; This paper retraces the state of the art, the recent development and the unsolved issues found in the composites manufacturing processes by resin transfer molding. We are revisiting the main technological but also scientific issues as the ones related to the identification of the preform permeability, the control and optimization issues for industrial applications, and in particular in the aeronautic ones.

International audience; Simulation of RTM processes is usually performed in 2D. Thus, only a mesh of the middle plane is needed with the associated degrees of freedom savings with respect to fully 3D modeling. However such modeling needs the definition of an equivalent in-plane permeability representing the ignored dimension (the thickness). The definition of such permeability is not a trivial task because each ply in the thickness direction can be anisotropic, being the principal anisotropy direction different from one ply to the neighbor plies. In this work we propose a novel fully 3D modeling whose computational cost is equivalent to a 2D solution. It allows addressing properly the equivalent in-plane permeability issue.

Colloque avec actes et comité de lecture. Internationale.; International audience; La simulation efficace des modèles thermiques non-linéaires entraine un certain nombre de verrous associés : (i) au caractère multi-échelle en espace et temps ; (ii) aux grandes dimensions spatiales et temporelles ; (iii) au caractère non linéaire et aux termes sources décrivant différentes cinétiques... Dans ce travail nous allons nous affranchir de ces difficultés en combinant une technique de réduction de modèles et une technique de séparation de variables.

The numerical simulation of composite forming permits to envisage the feasibility of a process without defect but also to know the directions of the reinforcements after shaping. These directions condition strongly the mechanical behaviour of the final textile composite structure. In addition, the angles between warp and weft yarns influence the permeability of the reinforcement and thus the filling of the resin in the case of a liquid moulding process. The forming of composite reinforcement can be made on a single ply or simultaneously on several plies. In this paper models for the single woven ply forming simulation are described. The semi-discrete approach is used for the simultaneous hemispherical forming of three layers., La simulation numérique de la mise en forme des composites permet d'analyser la faisabilité d'un procédé et ses défauts, mais aussi de déterminer les directions des renforts après la mise en forme. Ces directions conditionnement fortement le comportement mécanique de la pièce composite en service. La mise en forme peut concerner un seul pli ou bien plusieurs simultanément. Dans ce travail, les différentes approches actuellement utilisées pour la simulation de la mise en forme d'un pli de renfort tissé sont décrites. En particulier l'approche semi discrète est utilisée pour le formage hémisphérique simultané de trois plis.

In the digital era, users share their personal data with service providers to obtain some utility, e.g., access to high-quality services. Yet, the induced information flows raise privacy and integrity concerns. Consequently, cautious users may want to protect their privacy by minimizing the amount of information they disclose to curious service providers. Service providers are interested in verifying the integrity of the users' data to improve their services and obtain useful knowledge for their business. In this work, we present a generic solution to the trade-off between privacy, integrity, and utility, by achieving authenticity verification of data that has been encrypted for offloading to service providers. Based on lattice-based homomorphic encryption and commitments, as well as zero-knowledge proofs, our construction enables a service provider to process and reuse third-party signed data in a privacy-friendly manner with integrity guarantees. We evaluate our solution on different use cases such as smart-metering, disease susceptibility, and location-based activity tracking, thus showing its versatility. Our solution achieves broad generality, quantum-resistance, and relaxes some assumptions of state-of-the-art solutions without affecting performance.