LIMSI is organized into two main scientific departments: the department of Mechanics Energetics (ME), and the department of Human-Machine Communication (CHM). The former is composed of three research groups and corresponds approximately to 35 to 40% of our research force; the latter is structured into six research groups including a transversal thematic action. Many cross-department activities and collaborative research actions contribute to strengthening the scientific unity of the laboratory
Mechanics Energetics (ME) Department
Fluids Mechanics and Heat and Mass Transfer are key scientific disciplines at the heart of many crucial societal challenges in the domain of energy, transportation, and environment. Achieving more efficient, more reliable, more environmentally friendly means of converting or using energy, of transporting people and goods, requires a better identification of the corresponding technological bottlenecks and in turn a deeper knowledge of the involved physical mechanisms in all their intrinsic complexity and mutual interactions. It also necessitates a continuous progress in numerical modeling and simulation capabilities that are instrumental to mastering and optimizing the technological processes that stand at the heart of a progressive substitution of empirical know-how by a deterministic approach in the conception and design processes. Along these lines, research in the Mechanical Engineering department strives towards a better understanding of fluid and transfer phenomena, isolated or in interaction, and the development of efficient methodologies, numerical or experimental, to improve the predictive capabilities of models and their numerical simulations.
Achieving better predictive capabilities is also instrumental for the development of methodologies aimed at optimizing or controlling the events or processes in which these phenomena are involved.
Our research thus follows a dual strategy, proceeding either along a deductive process, each research topic progressing along its own logic, or along an inductive process, building upon concrete situations or configurations raised by our industrial or societal partners. We strive to keep a balance between these two approaches in order to reconcile both objectives of contributing to knowledge advancement while putting our skills at work for the benefit of society, the main specificity of engineering sciences.
Research in the mechanics-energetics department addresses a large variety of situations in fluid mechanics and energetics. This variety stems from the range of scales which are addressed, from micro-meters in nanoscale heat transfer to tens of meters in outer aerodynamics; from the range of speeds, corresponding to incompressible flows characterized by small Reynolds numbers up to supersonic flows; from the variety of the investigation methodologies, covering both numerical and experimental techniques; from the variety of numerical methods which are used or under development, finite volume, finite element, deterministic or stochastic spectral methods as well as reduced order methods; and from the variety of objectives, from pure knowledge advancement to proof of the concept demonstrators. This diversity of methodologies and goals, which constitutes our common global scientific expertise, is organized in three research groups, Unsteady Aerodynamics (AERO), Convection and Rotation (CORO), and Solid-Fluid Transfer (TSF), which present their research individually in detail in this scientific report. These three groups share or jointly develop a large number of methodologies, both numerical and experimental. To situate our research in a more general perspective, it can be described along two main themes, “aerodynamics” and “transfer and energetics”, which provide a global coherence to our scientific project.
The laboratory has long been recognized for its skills in the methodological developments related to numerical simulation and modeling which remain essential to scientific progress in the field of Fluid Mechanics and Energetics. Besides our work on numerical methodology developments, mainly devoted to High-Performance-Computing, we have recently increased efforts on the development of uncertainty quantification (UQ) methods. Our progress in numerical methods, capitalized in massively parallel software (SFEMaNS, BLUE, SUNFLUIDH), allows us to focus on multiphysics phenomena (e.g. MagnetoHydroDynamic (MHD) coupling in Liquid Metal Batteries as energy storage medium, liquid/vapor phase change in distillation processes, or reactive flows for combustion and plasmas) to better understand the phenomena involved. Aspects related to the coupling between conduction, convection, and radiation have also been studied for high Rayleigh number configurations, either to optimize electricity production processes from solar, thermodynamic, or photovoltaic energies (collaboration LGEP) or in a more the oretical context to define scaling laws in natural convection such as applications addressing the thermal performance of buildings. Significant progress has also been made in improving understanding of basic mechanisms of transition to turbulence for better energy efficiency of aerodynamic devices and renewable energy systems. Studying heat transfers at the solid/fluid interface at very low temperatures still remains a challenge for establishing new physical laws and experimentally studying energy dissipation in micro- or nano-devices. These activities were conducted through scientific collaborations thanks to institutional instruments provided in the context of Paris-Saclay.
Because of the wide spectrum of research activities at LIMSI, work is also conducted in collaboration between research groups. The LIMSI has for many years focused on the understanding of voice production, which requires a multidisciplinary approach combining fluid mechanics, acoustics, and fluid/deformable structure interactions. This activity, supported by the International Associated Laboratory (LIA) “Physics & Fluid Mechanics”, is located at the interface between the AERO and Audio and Acoustics (AA) groups. To analyze and understand the physical mechanisms responsible for unsteady and large scale motions in fluid flows, it is instrumental to develop methods and tools capable of extracting relevant information from huge amounts of data coming from either simulations or experiments, to visualize this information and interact with the representation, approaching real-time. This is carried out in the LidEx “Center for Data Sciences” in a collaborative work between the AERO, TLP, and VENISE groups. Bringing together our expertise in HMI and Mechanics-Energetics has also led to the development of original haptic devices based on heated/cooled air jet for additional modalities and more versatile interactions.
Human-Machine Communication (CHM) Department
Research in Human-Machine Communication has become increasingly important over recent decades. Two main aspects of LIMSI’s research may be highlighted: how to interact with the machine, for which more and more (hardware or software)
solutions are explored, and how to process human language, which is the most natural means to exchange information. These axes constitute the two main research fields of the Human-Machine Communication Department.
Concerning interaction, our research focuses both on hardware aspects, with the development of new devices such as large tactile surfaces, haptic interfaces, and audio and/or visual reality augmenting tools for immersive interactions; and on software aspects: people interact more and more often with artificial companions would they be virtual agents, robots or connected objects. In immersive CAVE like environment, users can also interact with life-sized virtual humans.. Studying human behavior in their interaction with artificial environments is also critical for making these interactions as natural, efficient and useful as possible. A particular field of interest, which gathers several groups, is the study and design of interaction devices and modalities in artistic contexts.
The main part of the information exchanged between humans is expressed by the means of language: either in the form of texts, for instance on a newspaper’s web site, or recordings, such as broadcast news or goal-oriented dialogues. But today’s information is becoming more and more “user-generated”, and as a result, arguably less structured; the analysis, indexation, and understanding of these various forms of “noisy” texts (forums, blogs, and micro-blogs) and audio recordings (spontaneous conversations, recorded classes and presentations, home-made videos) are less studied and pose new challenges which need to be addressed, while also providing invaluable raw material for the in-vivo analysis of actual language use. Other dimensions of language variation are also studied, alone or in combination: accentuated or emotive speech needs to be characterized, transcribed and indexed, texts and talks in foreign languages need to be translated, technical texts in specialized domains (e.g. bio-medical) need to be indexed or searched.
There are no clear-cut boundaries between all these studies. Despite the fact that the department is administratively organized in groups, many projects actually gather researchers from various groups.
Research in the Human-Machine Communication Department addresses the different facets of communication activities. This diversity begins with the medium used to communicate: from speech to writing, sound and music, hand or full body gestures, visual contact, or a mix of these different media. The device used to interact with the machine, or with other humans via the machine, is another source of diversity: from classical WIMP interfaces (window, mouse, and keyboard), to Virtual and Augmented Reality user interfaces or intelligent sensors in an I-room.
Most of our research is concerned with multidisciplinary aspects: ergonomics (studying human activity and system use), psychology (understanding and modelling human behavior and major cognitive functions with or for the design of virtual systems), physical and perceptual aspects of acoustics (for research concerning audio), linguistics (for automatic language processing) and para-linguistics (for research on social and affective dimension of spoken and multimodal interactions), or cognitive science (for perceptual studies in robotics). The corresponding research questions require specialists from different disciplines in order to be answered.
Some research fields are strongly related to societal aspects: designing virtual signers for the deaf and Sign Language community, developing special interactions to help autistic people to communicate, to be companions for aged people, or audio interfaces to help the blind be more autonomous. All these research fields have important societal impacts.
The development of these activities requires high performance computing architectures to run demanding algorithms under strong communication constraints such as real-time operation or with the treatment of huge amounts of data.
Research is primarily organized in six groups. Human Language Technology mainly concerns the Information, Written, and Signed Language group (ILES) for research on written and signed language, and the Spoken Language Processing group (TLP) for spoken language. All the work performed in these groups is strongly related to the activities of the Institute for Multimedia and Multimodal information (IMMI), an international CNRS laboratory hosted within our premises with the aim of developing and supporting LIMSI's collaborations with two German universities (RTWH in Aachen and KIT in Karlsruhe).
The specificities of audio, speech, and music are studied within the Audio & Acoustics group (AA), at the frontier between Human Language Technology and Interaction, while also interacting strongly with the Mechanics Energetics department. The Architectures and Models for Interaction group (AMI) deals with a variety of software or devices used to interact with humans. Cognition, Perception, Use group (CPU) conducts multidisciplinary research. Researchers in computer science, psychology, and ergonomics collaborate to address research questions concerning Human-Computer Interaction from a psychological point of view, and symmetrically use virtual interfaces to better understand human main cognitive functions such as perception, emotion and learning. VENISE (for Virtual & augmented ENvIronments for Simulation & Experiments) is the main research group in Virtual and Augmented Reality (V&AR). It develops its activities mainly on the EVE system (Evolutive Virtual Environment), an immersive, multi-sensorimotor and reconfigurable CAVE-like equipment designed, installed and actively supported by the group. VENISE is especially involved in the scientific and technological direction of the EquipEx DIGISCOPE, whose research focus is the collaborative interaction within high-performance visualization infrastructures.
The transverse action VIDA (for Virtuality, Interaction, Design & Art) is a collaborative theme dedicated to collaborations with professional artists. Researchers involved in this theme are also members of the different groups of the CHM department. These collaborations between groups are very fruitful and enable us to address the problem as a whole: taking into account both computer and human features for example.