CITI seminar – Alessandro Renzaglia (CITI) – 13/12 at 12:15

Speaker: Alessandro Renzaglia (CR Inria, Chroma)

Date: 13/12/2022

Time: 12h15

Place: Amphi Chappe/Lamarr

Title: Cooperative Exploration for 3D Reconstruction with Multiple Aerial Robots

Abstract: Autonomous exploration of unknown environments is a fundamental task in many robotics applications, such as map reconstruction, search and rescue operations, and inspection tasks. For this reason, it is a widely studied problem and several efficient approaches, for both single and multi-robot systems, have been proposed over the years. However, most of the existing solutions focus on 2D areas, while complex 3D environments still present several problems to be solved. In this talk, I will present some of the challenges of exploring 3D environments with a fleet of cooperating aerial vehicles and discuss different ways to quantify the expected new information contained in unexplored portions of the map and to take into account potential redundancies in the observations. Finally, I will briefly discuss the case, common in inspection tasks, where not all areas to explore are equally informative, but some regions have a higher priority and introduce a search dimension to the classical exploration task.

Bio: Alessandro Renzaglia is a research faculty member in the Chroma team at Inria Lyon / CITI Lab. He received his M.Sc. degree in Physics from the University of Rome La Sapienza, Italy, and his Ph.D. degree in Computer Science from the University of Grenoble, France. Successively he has been Postdoctoral Researcher with the Computer Science & Engineer Department at the University of Minnesota, Minneapolis, USA, and later with the Laboratory for Analysis and Architecture of Systems (LAAS), Toulouse, France, in the Robotics and Interactions group, before to join the Chroma team. His main research interests include multi-robot systems, path planning, and optimization.


CITI seminar – Anastasia Volkova (Univ. Nantes) and Florent de Dinechin (CITI) – 08/11 at 12:15

Title: Reconciling LTI filters and their implementations

Date and Place: 12h15 Tuesday 08/11/2022 in Amphi Hedy Lamarr/Chappe

Speaker: Anastasia Volkova (Univ. Nantes) and Florent de Dinechin (CITI)

 

Abstract: 

Linear time-invariant (LTI) filters are widely used in digital signal processing (DSP). Such filters can be designed out of a frequency specification (amplify some frequency bands, attenuate others), and a wide body of techniques exist to transform such a specification into an actual implementation consisting of additions and multiplications. The linearity property (the L of LTI) is essential at all the steps of this design process. However, actual implementations in hardware or software cannot be linear, due to the rounding of intermediate results: rounding is necessary to keep the result of a multiplication on the same number of bits as its inputs. But rounding is not linear, for instance the function that rounds a real to the nearest integer is not linear. So implementations of LTI filters are actually not linear, and this loss of linearity destroys in principle most mathematical foundations on which they are built. The workaround found by the DSP community so far is quite simple: ignore the problem, and hope for the best. Most of the times it works, and indeed your mobile phone is full of such filters. When it doesn’t, strange (and sometimes catastrophic) things will happen, and the community has devised all sorts of tricks and patches to evade them.

This talk will first present a gentle (but incomplete) introduction to LTI filters, their design, and the issue of non-linearity. It will then introduce a simple formalization that reconciles LTI filters and their
implementations. An additional benefit of this unified view is that filter design and implementation becomes a well-formed single global optimisation problem. The talk will conclude with the latest developments towards solving this problem.

Bio:

Anastasia Volkova est maître de conférences à l’Université de Nantes et membre de l’équipe OGRE au LS2N. Avant de commencer à Nantes en 2019, elle a été chercheuse postdoctorale à Intel San Diego, au Max-Plank Institute à Saabrucken, et à l’Inria à Lyon. Dans sa recherche elle se focalise sur des problématiques de précision finie dans des applications numériques, notamment l’analyse d’erreurs, mais aussi sur le compromis entre la précision et la performance. Elle s’intéresse en particulier à la conception et à l’implémentation logicielle ou matérielle des filtres LTI avec garantie de qualité  numérique, et à l’optimisation des ressources matérielles, e.g. sur FPGA.

Florent de Dinechin est professeur à l’INSA-Lyon après un doctorat de l’Université of Rennes-1 en 1997, un post-doctorat à Imperial College à Londres, et un premier poste à l’École Normale Supérieure de Lyon. Il s’intéresse à l’arithmétique des ordinateurs dans ses aspects mathématiques, matériels et logiciels, et particulièrement aux fonctions élémentaires, à la virgule flottante, et à l’arithmétique pour les FPGA.


PhD Defence: “Symmetric semi-discrete optimal transport for mesh interpolation”, Agathe Herrou, C5 room, Nautibus building, 20th of October 2022 at 1.30 PM

 

The defenseIt will take place on Thursday 20th october at 1:30 PM in the C5 room of the Nautibus building, Villeurbanne.

 

Title

Symmetric semi-discrete optimal transport for mesh interpolation

 

 

Abstract

This thesis aims to develop geometric methods to approximate displacement interpolation, derived from optimal transport. Optimal transport is a mathematical theory modeling movements of matter under a cost minimization constraint, with many applications in physics, computer graphics and geometry. The minimum displacement cost between two distributions defines a distance, which itself is at the origin of displacement interpolation. This interpolation may under certain conditions present discontinuities, that the discretized approximations of the optimal transport do not always successfully capture. The work of this thesis aims to develop an approximation that captures these discontinuities well. Our method relies on semi-discrete optimal transport, where only one of the distributions is discretized, thus accurately capturing the discontinuities of the distribution that remains continuous. The transport plans thus obtained partition the continuous distribution into cells associated with the samples of the discretization. A semi-discrete optimal transport plan can thus be assimilated to a power diagram made up of these cells. This variant of optimal transport however has the disadvantage of breaking the symmetry between the two distributions. We start by formalizing our problem as the search for a pair of transport plans coupled through the barycenters of their cells. We then present an algorithm for calculating these coupled transport plans. This first algorithm is based on a classical alternating algorithm scheme, successively computing the transport plans and the barycenters of their cells until convergence. The results obtained from this algorithm allow to interpolate between the initial distributions while maintaining a satisfactory precision, in particular when it comes to discontinuities, including when the discretization of the distributions is done with relatively few points. We then present our exploration of optimization methods for solving the same problem. These methods express the constraints of our problem as a critical point of a functional, and aim to reach these points using algorithms such as Newton’s method. However, this approach did not yield conclusive results, as the functions involved were too noisy to lend themselves well to optimization algorithms.

Keywords: Optimal transport, Interpolation, Optimization, Algorithmic geometry

Jury

    • Julie Delon (reviewer), Université Paris Cité
    • Boris Thibert (reviewer), Université Grenoble Alpes
    • Dominique Attali (examiner), CNRS/Université Grenoble Alpes
    • Filippo Santambrogio (examiner), Université Lyon 1
    • Nicolas Bonneel (advisor), CNRS/Université Lyon 1
    • Julie Digne (co-advisor), CNRS/Université Lyon 1
    • Bruno Lévy (co-advisor), Inria Nancy Grand Est

CITI seminar – Petre-Daniel Matasaru (“Gheorghe Asachi” Technical University Iasi) – 01/09 at 10:00

Title: Research Activities and Interests for Potentially Future Collaborations

Date and Place: 10h Thursday 1/9/2022 in TD-C

Speaker: Petre-Daniel Matasaru (“Gheorghe Asachi” Technical University Iasi)

Hosts: Prof. Oana Iova (AGORA) 

 

Abstract: 

The presentation is composed of 4 major axis: the City of Iasi, an overview of the Research at TUIASI, a detailed look at the Research and PhD Programs at ETTI and the Researcher’s personal research activities and interests with a focus on most relevant work.
A kit that was developed through a collaboration between the private sector (Bitron Electronics Engineering SRL, the Iasi (Romania) branch of Bitron Group, a worldwide Italian company) and the academic staff from ETTI will be presented. The development team included several disciplines such as hardware, software, PCB layout, mechanics and project-management and the resulting IoT kit can be used by researchers and students to develop and test various applications within the Internet of Things framework, highlighting the components of this kit and its capabilities for potential further development in the automotive area. Different sensors acquire information from the physical environment and data is being sent to a remote communication node that centralizes data traffic in a wireless sensor network. Authors developed a C program for an encryption application that runs on the board in order to secure the values obtained from the sensors and sent to the central node by the wireless adapter.

Another relevant paper called “Experimental and theoretical investigations of a plasma fireball dynamics” refers to modelling in the frame of the scale relativity model of the periodic current bursts observed in the dynamic current-voltage characteristic of a probe in the presence of a plasma fireball in dynamic state, based on both the fractal space-time concept and the generalization of Einstein’s principle of relativity to scale transformations. The bursts appear in the probe characteristic when a certain relation exists between the fireball dynamics frequency and the frequency of the probe voltage sweep. The double layer dynamics is described by a set of time-dependent Schrödinger-type equations and the self-structuring is given by means of the negative differential resistance. The obtained experimental and theoretical results are proven to be in very good agreement.”

 

Bio:

Born in Alexandria, Romania, studying electronics and telecommunications engineering at the Technical University of Iasi, graduating BSc in 2002 with a diploma thesis done at the Technische Universitat Darmstadt (TUD), Germany, due to an Erasmus mobility. MSc in 2003 and PhD with the thesis “Probabilistic methods used in the simulation of radio frequency structures” in 2011. Assistant-professor and Lecturer starting since 2003, at the Telecommunications Department of ETTI Iasi. Member of the Mod-Sim-Nano Research Group, Erasmus+ Coordinator at ETTI and member of the University Senate. Didactic mobility at Universita degli Studi Palermo, Italy (UNIPA) in 2019.


CITI seminar – Jérôme Nika (Ircam) – 4/7 at 10:00

Title: Technologies génératives pour la création musicale : composer à l’échelle du comportement ou de la narration.

Date and Place: 10h Monday 4/7/2022 in TD-C

Speaker: Jérôme Nika (Ircam)

Hosts: Romain Michon and Tanguy Risset (Emeraude project-team)

 

Abstract: 

Une machine sera-t-elle bientôt capable de remplacer l’humain dans la création musicale ? Pour toute une partie des artisans de l’intelligence artificielle appliquée à la musique, artistes comme scientifiques, il est difficile de répondre à cette question récurrente car ce n’est pas celle qui se pose. En effet, si on “apprend” la musique à des ordinateurs dotés d’une “mémoire” musicale inspirée de la cognition humaine, l’enjeu réside précisément dans le fait de partir de ces modèles pour explorer la production d’une musique nouvelle plutôt que la reproduction d’une musique crédible.

Les recherches associées au sein de l’équipe Représentations Musicales de l’Ircam s’articulent autour de la notion de « mémoire musicale » : son apprentissage, sa modélisation, et sa mobilisation dans un contexte créatif. Elles ont donné naissance à DYCI2lib, une librairie d’agents génératifs pour la performance et la composition musicale combinant les approches libres, planifiées et spécifiées, et réactives de la génération à partir d’un corpus. Ces travaux ont été mis en oeuvres dans le cadre de nombreuses collaborations artistiques et productions musicales, notamment dans le jazz et les musiques improvisées (Steve Lehman, Bernard Lubat, Benoît Delbecq, Rémi Fox, Orchestre National de Jazz), la musique contemporaine (Pascal Dusapin, Ensemble Modern, Alexandros Markeas, Marta Gentilucci), et l’art contemporain (Le Fresnoy – Studio National des Arts Contemporains, Vir Andres Hera) La présentation des pratiques musicales permises par ces instruments génératifs au service de la créativité humaine sera illustrée par des extraits de ces productions récentes.

 

Bio:

Jérôme Nika est chercheur en technologies génératives pour la création musicale, réalisateur en informatique musicale, et musicien. Diplômé des écoles ENSTA ParisTech et Télécom ParisTech ainsi que du master ATIAM (Acoustique, Traitement du signal, Informatique, Appliqués à la Musique – Sorbonne Université / Ircam / Télécom ParisTech), il a également étudié la composition musicale. Il s’est spécialisé dans l’application de l’informatique et du traitement du signal à la création numérique et à la musique à travers un doctorat (« Prix Jeune Chercheur Science/Musique 2015 », « Prix Jeune Chercheur 2016 », Association Française d’Informatique Musicale) puis en tant que chercheur à l’Ircam (Institut de Recherche et Coordination Acoustique/Musique).

En 2019, il entre au Fresnoy – Studio National des Arts Contemporain en tant que chercheur invité. Cette même année, il travaille sur 3 projets : le projet évolutif Lullaby Experience, du compositeur Pascal Dusapin et deux projets de musique improvisée : Silver Lake Studies en duo avec Steve Lehman et C’est pour ça en duo avec Rémi Fox (projet lauréat de l’aide DICRéAM du CNC pour 2020) .

En 2020, il devient chercheur permanent dans l’équipe Représentations Musicales de l’Ircam où il développe des instruments logiciels en interaction avec des musiciens experts. Plus de 60 concerts et performances artistiques ont mis ces outils en jeu depuis 2016 (Onassis Center, Athènes, Grèce; Ars Electronica Festival, Linz, Autriche; Frankfurter Positionen festival, Frankfurt; Annenberg Center, Philadelphia, USA; Centre Pompidou, Collège de France, LeCentquatre, Paris, France; Montreux Jazz festival, etc.). La dernière production en date, pour laquelle il crée l’électronique générative est le concert “Ex Machina”, une collaboration entre Steve Lehman, Frédéric Maurin, et l’Orchestre National de Jazz, créé le 11 février 2022 à la Maison de la Radio et diffusé sur France Musique.

 


PhD Defence: “User Association in Flexible and Agile Mobile Networks”, Romain PUJOL, amphi, Chappe Building, 6th of July 2022 at 2 PM

 

The defense will take place on Wenesday, July 6 at 2 pm, in the amphitheatre of the Telecommunications Department (Claude Chappe building), INSA Lyon, Villeurbanne.

The presentation will be available online: https://insa-lyon-fr.zoom.us/j/91599742042

 

Title

User Association in Flexible and Agile Mobile Networks

 

 

Abstract

The user association process in cellular networks consists in choosing the base station with which the user equipment will negotiate radio resources. The current association is based on measuring the signal strength received by the user equipment from each of the base stations. The association must now deal with the diversification of user application needs and the growing heterogeneity of cellular networks.

 

In this thesis, we show experimentally that the current association process has reached its limits, that it is agnostic of the configurations of the base stations and that it does not allow to control the throughput that the user equipment will obtain following the association. We set up for the realization of our measurements an experimental platform based on the software suite of cellular networks srsRAN and software radios USRP NI-2091.

 

We propose in this thesis a new metric to be used during the association process. This metric, broadcasted by the base station, is a load information which in our case will be represented by the number of user equipment connected to the base station. We also discuss other metrics that can be used as load information. We show, once again experimentally, by modifying the source code of the srsRAN software suite ourselves, that if the user equipment takes this load information into account in the association process, the association decision is improved in 22% of cases.

 

Jury

  • Tara ALI YAHIA, Associate Professor HDR, Université Paris Saclay (Reviewer)
  • Vania CONAN, HDR, Thales (Reviewer)
  • Véronique VEQUE, Professor, Université Paris Saclay (Examiner)
  • Frédéric LAUNAY, Associate Professor, IUT de Poitiers (Examiner)
  • Stéphane FRENOT, Professor, INSA-Lyon (Examiner)
  • André-Luc BEYLOT, Professor, ENSEEIHT (Examiner)
  • Fabrice VALOIS, Professor, INSA-Lyon (Thesis supervisor)
  • Razvan STANICA, Associate Professor HDR, INSA-Lyon (Thesis co-supervisor)

PhD Defence: “Activity Models and Bayesian Estimation Algorithms for Wireless Grant-Free Random Access”, Lélio CHETOT, amphi, Chappe Building, 7th of July 2022 at 2 PM

 

The defense will take place on Thursday, July 7 at 2 pm, in the amphitheatre of the Telecommunications Department (Claude Chappe building), INSA Lyon, Villeurbanne.

The presentation will be available online: https://youtu.be/hX3t9pKPcoc

 

Title

Activity Models and Bayesian Estimation Algorithms for Wireless Grant-Free Random Access

 

 

Abstract

The new 5G’s wireless networks have recently started to be deployed all around the world. With them, a large spectrum of services are about to emerge, resulting in new stringent requirements so that 5G targets performance exceed that of 4G by a factor of 10. The services are centered around the use cases of enhanced mobile broadband (eMBB), ultra reliable and low-latency communication (uRLLC) and massive machine-type communication (mMTC) where each of which has required the ongoing development of key new technologies. Many of these technologies will also play an important role in the emergence of 6G.

In this thesis, the focus is on grant-free RA (GFRA) as an enabler of uRLLC and mMTC. GFRA is a new protocol introduced in 5G new radio (5G-NR) for reducing the data overhead of the random access (RA) procedure. This results in a significant reduction in the latencies of the user equipments (UEs) access to a connected medium via an access point (AP). Achieving efficient GFRA is of key importance for many 5G applications, e.g. for large scale internet of things (IoT) wireless networks. The study of new non-orthogonal multiple access (NOMA) signal processing techniques is then considered. Using tools from the theory of compressed sensing (CS), and particularly from Bayesian CS, new algorithms within the family of approximate message passing (AMP) are developed to address the joint active user detection and channel estimation (AUDaCE) problem. The active user detection is crucial to properly identify transmitting UEs within the context of large-scale dense network; the channel estimation is equally important so that an AP can reliably transmit back data to the detected UEs.

In this thesis, in contrast to existing work on this topic, the AUDaCE is studied for wireless networks where the activity of the UEs is assumed to be correlated, as is typical for many large-scale dense networks. To this end, two new activity models are introduced. The first one assumes that the activity of the UEs in the network can be modeled via group-homogeneous activity (GHomA) where devices in the same group have common pairwise correlations and marginal activity probabilities. The second model accounts for more general dependence structure via group-heterogeneous activity (GHetA). Novel approximate message passing algorithms within the hybrid GAMP (HGAMP) framework are developed for each of the models. With the aid of latent variables associated to each group for modeling the activity probabilities of the UEs, the GHomA-HGAMP algorithm can perform AUDaCE for GFRA leveraging such a group homogeneity. When the activity is heterogenous, i.e. each UE is associated with a latent variable modeling its activity probability correlated with the other variables, it is possible to develop GHetA-HGAMP using the copula theory.

Extensive numerical studies are performed, which highlight significant performance improvements of GHomA-HGAMP and GHetA-HGAMP over existing algorithms (modified generalized AMP (GAMP) and group-sparse HGAMP (GS-HGAMP)), which do not properly account for correlation in activity. In particular, the channel estimation and active user detection capability are enhanced in many scenarios with up to a 4dB improvement with twice less user errors.

As a whole, this thesis provides a systematic approach to AUDaCE for wireless networks with correlated activities using tools from Bayesian CS. We then conclude by showing how it could be used for multi-carrier orthogonal frequency division multiplexing (OFDM) scenarios with possible extensions for grant-free (GF) data transmission leveraging joint data recovery, active user detection and channel estimation (DrAUDaCE).

 

Jury

  • Catherine DOUILLARD, Professor @ IMT Atlantique, Reviewer
  • Dejan VUKOBRATOVIC, Professor @ Univ. Novi Sad, Reviewer
  • Aline ROUMI, Senior Research Scientist @ INRIA Rennes, Examiner
  • Philippe CIBLAT, Professor @ Telecom Paris, Examiner
  • Cedomir STEFANOVIC, Professor @ Univ. Aalborg, Examiner
  • Jean-Marie GORCE, Professor @ INSA Lyon, Director
  • Malcolm EGAN, Research Scientist @ INRIA Lyon, Supervisore

PhD Defence: “Large-scale Automatic Learning of Autonomous Agent Behavior with Structured Deep Reinforcement Learning”, Edward Beeching, amphi, Chappe Building, 3rd of May 2022 at 10:00 AM

 

The defense willtake place on Tuesday, May 3 at 10:00 am, in the amphitheatre of the Telecommunications Department (Claude Chappe building), INSA Lyon, Villeurbanne.

The presentation will be available on Youtube at the following link: https://youtu.be/k7dh-thqbSk

 

Title

Large-scale Automatic Learning of Autonomous Agent Behavior with Structured Deep Reinforcement Learning

 

 

Abstract

Autonomous robotic agents have begun to impact many aspects of our society,with application in automated logistics, autonomous hospital porters, manufacturing and household assistants. The objective of this thesis is to explore Deep Reinforcement Learning approaches to planning and navigation in large and unknown 3D environments. In particular, we focus on tasks that require exploration and memory in simulated environments. An additional requirement is that learned policies should generalize to unseen map instances. Our long-term objective is the transfer of a learned policy to a real-world robotic system. Reinforcement learning algorithms learn by interaction. By acting with the objective of accumulating a task-based reward, an Embodied AI agent must learn to discover relevant semantic cues such as object recognition and obstacle avoidance, if these skills are pertinent to the task at hand. This thesis introduces the field of Structured Deep Reinforcement Learning and then describes 5 contributions that were published during the PhD.

We start by creating a set of challenging memory-based tasks whose performance is benchmarked with an unstructured memory-based agent. We then demonstrate how the incorporation of structure in the form of a learned metric map, differentiable inverse projective geometry and self-attention mechanisms; augments the unstructured agent, improving its performance and allowing us to interpret the agent’s reasoning process.

We then move from complex tasks in visually simple environments, to more challenging environments with photo-realistic observations, extracted from scans of real-world buildings. In this work we demonstrate that augmenting such an agent with a topological map can improve its navigation performance. We achieve this by learning a neural approximation of a classical path planning algorithm, which can be utilized on graphs with uncertain connectivity.

From work undertaken over the course of a 4-month internship at the research and development department of Ubisoft, we demonstrate that structured methods can also be used for navigation and planning in challenging video game environments. Where we couple a lower level neural policy with a classical planning algorithm to improve long-distance planning and navigation performance in vast environments of 1km×1km. We release an open-source version of the environment as a benchmark for navigation in large-scale environments.

Finally, we develop an open-source Deep Reinforcement Learning interface for the Godot Game Engine. Allowing for the construction of complex virtual worlds and the learning of agent behaviors with a suite of state-of-the-art algorithms. We release the tool with a permissive open-source (MIT) license, to aid researchers in their pursuit of complex embodied AI agents.

 

 

Jury

    • Mme. Elisa Fromont – Université de Rennes 1 – Rapporteur
    • M. David Filliat – ENSTA Paris – Rapporteur
    • M. Cédric Démonceaux – Université de Bourgogne – Examinateur
    • M. Karteek Alahari – INRIA Grenoble – Examinateur
    • Mme. Christine Solnon – INSA-Lyon – Examinateur
    • M. Olivier Simonin – INSA-Lyon – Directeur de thèse
    • M. Jilles Dibangoye – INSA-Lyon – Co-encadrant de thèse
    • M. Christian Wolf – Naver Labs Europe – Co-directeur de thèse