ERC FUNDED PROJECTS

"Combustion is essential to the world’s energy generation and transport needs, and will remain so for the foreseeable future. Mitigating its impact on the climate and human health, by reducing its associated emissions, is thus a priority. One significant challenge for gas-turbine combustion is combustion instability, which is currently inhibiting reductions in NOx emissions (these damage human health via a deterioration in air quality). Combustion instability is caused by a two-way coupling between unsteady combustion and acoustic waves - the large pressure oscillations that result can cause substantial mechanical damage. Currently, the lack of fast, accurate modelling tools for combustion instability, and the lack of reliable ways of suppressing it are severely hindering reductions in NOx emissions. This proposal aims to make step improvements in both fast, accurate modelling of combustion instability, and in developing reliable active control strategies for its suppression. It will achieve this by coupling low order combustor models (these are fast, simplified models for simulating combustion instability) with advances in analytical modelling, CFD simulation, reduced order modelling and control theory tools. In particular: * important advances in accurately incorporating the effect of entropy waves (temperature variations resulting from unsteady combustion) and non-linear flame models will be made; * new active control strategies for achieving reliable suppression of combustion instability, including from within limit cycle oscillations, will be developed; * an open-source low order combustor modelling tool will be developed and widely disseminated, opening access to researchers worldwide and improving communications between the fields of thermoacoustics and control theory. Thus the proposal aims to use analytical and computational methods to contribute to achieving low NOx gas-turbine combustion, without the penalty of damaging combustion instability."

1 489 309 €

Start date: 2013-01-01, End date: 2017-12-31

"Computer vision is concerned with the automated interpretation of images and video streams. Today's research is (mostly) aimed at answering queries such as ""Is this a picture of a dog?"", (classification) or sometimes ""Find the dog in this photo"" (detection). While categorisation and detection are useful for many tasks, inferring correct class labels is not the final answer to visual recognition. The categories and locations of objects do not provide direct understanding of their function i.e., how things work, what they can be used for, or how they can act and react. Such an understanding, however, would be highly desirable to answer currently unsolvable queries such as ""Am I in danger?"" or ""What can happen in this scene?"". Solving such queries is the aim of this proposal. My goal is to uncover the functional properties of objects and the purpose of actions by addressing visual recognition from a different and yet unexplored perspective. The main novelty of this proposal is to leverage observations of people, i.e., their actions and interactions to automatically learn the use, the purpose and the function of objects and scenes from visual data. The project is timely as it builds upon the two key recent technological advances: (a) the immense progress in visual recognition of objects, scenes and human actions achieved in the last ten years, as well as (b) the emergence of a massive amount of public image and video data now available to train visual models. ACTIVIA addresses fundamental research issues in automated interpretation of dynamic visual scenes, but its results are expected to serve as a basis for ground-breaking technological advances in practical applications. The recognition of functional properties and intentions as explored in this project will directly support high-impact applications such as detection of abnormal events, which are likely to revolutionise today's approaches to crime protection, hazard prevention, elderly care, and many others."

1 497 420 €

Start date: 2013-01-01, End date: 2018-12-31

The objective of this proposal is to bring together a local team of young researchers who will work closely with international collaborators to advance the state of the art of Algorithmic Game Theory and open new venues of research at the interface of Computer Science, Game Theory, and Economics. The proposal consists mainly of three intertwined research strands: algorithmic mechanism design, price of anarchy, and online algorithms. Specifically, we will attempt to resolve some outstanding open problems in algorithmic mechanism design: characterizing the incentive compatible mechanisms for important domains, such as the domain of combinatorial auctions, and resolving the approximation ratio of mechanisms for scheduling unrelated machines. More generally, we will study centralized and distributed algorithms whose inputs are controlled by selfish agents that are interested in the outcome of the computation. We will investigate new notions of mechanisms with strong truthfulness and limited susceptibility to externalities that can facilitate modular design of mechanisms of complex domains. We will expand the current research on the price of anarchy to time-dependent games where the players can select not only how to act but also when to act. We also plan to resolve outstanding questions on the price of stability and to build a robust approach to these questions, similar to smooth analysis. For repeated games, we will investigate convergence of simple strategies (e.g., fictitious play), online fairness, and strategic considerations (e.g., metagames). More generally, our aim is to find a productive formulation of playing unknown games by drawing on the fields of online algorithms and machine learning.

2 461 000 €

Start date: 2013-04-01, End date: 2019-03-31

Back pain affects eight out of ten adults during their lifetime. It a huge economic burden on society, estimated to cost as much as 1-2% of gross national product in several European countries. Treatments for back pain have lower levels of success and are not as technologically mature as those for other musculoskeletal disorders such as hip and knee replacement. This application proposes to tackle one of the major barriers to the development of better surgical treatments for back pain. At present, new spinal devices are commonly assessed in isolation in the laboratory under standardised conditions that do not represent the variation across the patient population. Consequently many interventions have failed during clinical trials or have proved to have poor long term success rates. Using a combination of computational and experimental models, a new testing methodology will be developed that will enable the variation between patients to be simulated for the first time. This will enable spinal implants and therapies to be more robustly evaluated across a virtual patient population prior to clinical trial. The tools developed will be used in collaboration with clinicians and basic scientists to develop and, crucially, optimise new treatments that reduce back pain whilst preserving the unique functions of the spine. If successful, this approach could be translated to evaluate and optimise emerging minimally invasive treatments in other joints such as the hip and knee. Research in the spine could then, for the first time, lead rather than follow that undertaken in other branches of orthopaedics.

1 498 777 €

Start date: 2012-12-01, End date: 2018-11-30

In the last decade, solar and photovoltaic (PV) technologies have emerged as a potentially major technology for power generation in the world. So far the PV field has been dominated by silicon devices, even though this technology is still expensive.Dye-sensitized solar cells (DSC) are an important type of thin-film photovoltaics due to their potential for low-cost fabrication and versatile applications, and because their aesthetic appearance, semi-transparency and different color possibilities.This advantageous characteristic makes DSC the first choice for building integrated photovoltaics.Despite their great potential, DSCs for building applications are still not available at commercial level. However, to bring DSCs to a marketable product several developments are still needed and the present project targets to give relevant answers to three key limitations: encapsulation, glass substrate enhanced electrical conductivity and more efficient and low-cost raw-materials. Recently, the proponent successfully addressed the hermetic devices sealing by developing a laser-assisted glass sealing procedure.Thus, BI-DSC proposal envisages the development of DSC modules 30x30cm2, containing four individual cells, and their incorporation in a 1m2 double glass sheet arrangement for BIPV with an energy efficiency of at least 9% and a lifetime of 20 years. Additionally, aiming at enhanced efficiency of the final device and decreased total costs of DSCs manufacturing, new materials will be also pursued. The following inner-components were identified as critical: carbon-based counter-electrode; carbon quantum-dots and hierarchically TiO2 photoelectrode. It is then clear that this project is divided into two research though parallel directions: a fundamental research line, contributing to the development of the new generation DSC technology; while a more applied research line targets the development of a DSC functional module that can be used to pave the way for its industrialization.

1 989 300 €

Start date: 2013-03-01, End date: 2018-08-31

"Photovoltaics (PV) based on Silicon (Si) semiconductors is one the most growing technology in the World for renewable, sustainable, non-polluting, widely available clean energy sources. Theoretical and applied research aims at increasing the conversion efficiency of PV modules and their lifetime. The Si crystalline microstructure has an important role on both issues. Grain boundaries introduce additional resistance and reduce the conversion efficiency. Moreover, they are prone to microcracking, thus influencing the lifetime. At present, the existing standard qualification tests are not sufficient to provide a quantitative definition of lifetime, since all the possible failure mechanisms are not accounted for. In this proposal, an innovative computational approach to design and durability assessment of PV modules is put forward. The aim is to complement real tests by virtual (numerical) simulations. To achieve a predictive stage, a challenging multi-field (multi-physics) computational approach is proposed, coupling the nonlinear elastic field, the thermal field and the electric field. To model real PV modules, an adaptive multi-scale and multi-field strategy will be proposed by introducing error indicators based on the gradients of the involved fields. This numerical approach will be applied to determine the upper bound to the probability of failure of the system. This statistical assessment will involve an optimization analysis that will be efficiently handled by a Mathematica-based hybrid symbolic-numerical framework. Standard and non-standard experimental testing on Si cells and PV modules will also be performed to complement and validate the numerical approach. The new methodology based on the challenging integration of advanced physical and mathematical modelling, innovative computational methods and non-standard experimental techniques is expected to have a significant impact on the design, qualification and lifetime assessment of complex PV systems."

1 483 980 €

Start date: 2012-12-01, End date: 2017-11-30

The overall goal of the project is to contribute to the consolidation of the nascent and revolutionary philosophy of “Materials by Design” by resorting to the enormous power provided by the nowadays-available computational techniques. Limitations of current procedures for developing material-based innovative technologies in engineering, are often made manifest; many times only a catalog, or a data basis, of materials is available and these new technologies have to adapt to them, in the same way that the users of ready-to-wear have to take from the shop the costume that fits them better, but not the one that fits them properly. This constitutes an enormous limitation for the intended goals and scope. Certainly, availability of materials specifically designed by goal-oriented methods could eradicate that limitation, but this purpose faces the bounds of experimental procedures of material design, commonly based on trial and error procedures. Computational mechanics, with the emerging Computational Materials Design (CMD) research field, has much to offer in this respect. The increasing power of the new computer processors and, most importantly, development of new methods and strategies of computational simulation, opens new ways to face the problem. The project intends breaking through the barriers that presently hinder the development and application of computational materials design, by means of the synergic exploration and development of three supplementary families of methods: 1) computational multiscale material modeling (CMM) based on the bottom-up, one-way coupled, description of the material structure in different representative scales, 2) development of a new generation of high performance reduced-order-modeling techniques (HP-ROM), in order to bring down the associated computational costs to affordable levels, and 3) new computational strategies and methods for the optimal design of the material meso/micro structure arrangement and topology (MATO) .

2 372 973 €

Start date: 2013-02-01, End date: 2018-01-31

"The last few years have seen a growing interest for computational cell mechanics. This field encompasses different scales ranging from individual monomers, cytoskeleton constituents, up to the full cell. Its focus, fueled by the development of interdisciplinary collaborative efforts between engineering, computer science and biology, until recently relatively isolated, has allowed for important breakthroughs in biomedicine, bioengineering or even neurology. However, the natural “knowledge barrier” between fields often leads to the use of one numerical tool for one bioengineering application with a limited understanding of either the tool or the field of application itself. Few groups, to date, have the knowledge and expertise to properly avoid both pits. Within the computational mechanics realm, new methods aim at bridging scale and modeling techniques ranging from density functional theory up to continuum modeling on very large scale parallel supercomputers. To the best of the knowledge of the author, a thorough and comprehensive research campaign aiming at bridging scales from proteins to the cell level while including its interaction with its surrounding media/stimulus is yet to be done. Among all cells, neurons are at the heart of tremendous medical challenges (TBI, Alzheimer, etc.). In nearly all of these challenges, the intrinsic coupling between mechanical and chemical mechanisms in neuron is of drastic relevance. I thus propose here the development of a neuron model constituted of length-scale dedicated numerical techniques, adequately bridged together. As an illustration of its usability, the model will be used for two specific applications: neurite growth and electrical-chemical-mechanical coupling in neurons. This multiscale computational framework will ultimately be made available to the bio- medical community to enhance their knowledge on neuron deformation, growth, electrosignaling and thus, Alzheimer’s disease, cancer or TBI."

1 128 960 €

Start date: 2013-05-01, End date: 2018-04-30

"The goal of CV-SUPER is to create the technology to perform dynamic visual scene understanding from the perspective of a moving human observer. Briefly stated, we want to enable computers to see and understand what humans see when they navigate their way through busy inner-city locations. Our target scenario is dynamic visual scene understanding in public spaces, such as pedestrian zones, shopping malls, or other locations primarily designed for humans. CV-SUPER will develop computer vision algorithms that can observe the people populating those spaces, interpret and understand their actions and their interactions with other people and inanimate objects, and from this understanding derive predictions of their future behaviors within the next few seconds. In addition, we will develop methods to infer semantic properties of the observed environment and learn to recognize how those affect people’s actions. Supporting those tasks, we will develop a novel design of an object recognition system that scales up to potentially hundreds of categories. Finally, we will bind all those components together in a dynamic 3D world model, showing the world’s current state and facilitating predictions how this state will most likely change within the next few seconds. These are crucial capabilities for the creation of technical systems that may one day assist humans in their daily lives within such busy spaces, e.g., in the form of personal assistance devices for elderly or visually impaired people or in the form of future generations of mobile service robots and intelligent vehicles."

1 499 960 €

Start date: 2012-11-01, End date: 2017-10-31

We propose to radically extend the frontiers of two major themes in computing, parallelism and dynamism, and develop a novel paradigm of computing: dynamic-parallelism. To this end, we will follow two lines of research. First, we will develop techniques for extracting efficiency and high performance from parallel programs written in high-level programming languages. Second, we will develop the dynamic-parallelism model, where computations can respond to a wide variety of dynamic changes to their data automatically and efficiently, by developing novel abstractions (calculi), high-level programming-language constructs, and compilation techniques. The research will culminate in a language that extends the C programming language with support for parallel and dynamic-parallel programming. The proposal is motivated by urgent needs driven by the advent of multicore chips, which is making parallelism mainstream, and the increasing ubiquity of software, which requires applications to operate on highly dynamic data. These advances demand parallel and highly dynamic software, which remains too difficult and labor intensive to develop. The urgency is further underlined by the increasing data and problem sizes---online data grows exponentially, doubling every few years---that require similarly powerful advances in performance. The proposal will achieve profound impact by dramatically simplifying the development of high-performing dynamic and dynamic-parallel software. As a result, programmer productivity and software quality including correctness, reliability, performance, and resource (e.g., time and energy) consumption will improve significantly. The proposal will not only open new research opportunities in parallel computing, programming languages, and compilers, but also in other fields where parallel and dynamic problems abound, e.g., algorithms, computational biology, geometry, graphics, machine learning, and software systems.

1 076 570 €

Start date: 2013-06-01, End date: 2018-05-31

"The notion of meaning is central to many areas of Computer Science, Artificial Intelligence (AI), Linguistics, Philosophy, and Cognitive Science. A formal account of the meaning of natural language utterances is crucial to AI, since an understanding of natural language is at the heart of much intelligent behaviour. More specifically, Natural Language Processing (NLP) --- the branch of AI concerned with the automatic processing, analysis and generation of text --- requires a model of meaning for many of its tasks and applications. There have been two main approaches to modelling the meaning of language in NLP. The first, the ``compositional"" approach, is based on classical ideas from Philosophy and Mathematical Logic, and includes formal accounts of how the meaning of a sentence can be determined from the relations of words in a sentence. The second, more recent approach focuses on the meanings of the words themselves. This is the ``distributional"" approach to lexical semantics and is based on the idea that the meanings of words can be determined by considering the contexts in which words appear in text. The ambitious idea in this proposal is to exploit the strengths of the two approaches, by developing a unified model of distributional and compositional semantics, and exploiting it for NLP tasks and applications. The aim is to make the following fundamental contributions: 1. advance the theoretical study of meaning in Linguistics, Computer Science and AI; 2. develop new meaning-sensitive approaches to NLP applications which can be robustly applied to naturally occurring text. The claim is that language technology based on ``shallow"" approaches is reaching its performance limit, and the next generation of language technology requires a more sophisticated, but robust, model of meaning, which this project will provide."

1 087 930 €

Start date: 2012-09-01, End date: 2017-08-31

Flows of complex fluids, such as many biological fluids and most synthetic fluids, are common in our daily life and are very important from an industrial perspective. Because of their inherent nonlinearity, the flow of complex viscoelastic fluids often leads to counterintuitive and complex behaviour and, above critical conditions, can prompt flow instabilities even under low Reynolds number conditions which are entirely absent in the corresponding Newtonian fluid flows. The primary goal of this project is to substantially expand the frontiers of our current knowledge regarding the mechanisms that lead to the development of such purely-elastic flow instabilities, and ultimately to understand the transition to so-called “elastic turbulence”, a turbulent-like phenomenon which can arise even under inertialess flow conditions. This is an extremely challenging problem, and to significantly advance our knowledge in such important flows these instabilities will be investigated in a combined manner encompassing experiments, theory and numerical simulations. Such a holistic approach will enable us to understand the underlying mechanisms of those instabilities and to develop accurate criteria for their prediction far in advance of what we could achieve with either approach separately. A deep understanding of the mechanisms generating elastic instabilities and subsequent transition to elastic turbulence is crucial from a fundamental point of view and for many important practical applications involving engineered complex fluids, such as the design of microfluidic mixers for efficient operation under inertialess flow conditions, or the development of highly efficient micron-sized energy management and mass transfer systems. This research proposal will create a solid basis for the establishment of an internationally-leading research group led by the PI studying flow instabilities and elastic turbulence in complex fluid flows.

994 110 €

Start date: 2012-10-01, End date: 2018-01-31

Project FLEXABLE lies in the field of 3D Computer Vision, which seeks to recover depth or the 3D shape of the observed environment from images. One of the most successful and mature techniques in 3D Computer Vision is Shape-from-Motion which is based on the well-established theory of Multiple-View Geometry. This uses multiple images and assumes that the environment is rigid. The world is however made of objects which move and undergo deformations. Researchers have tried to extend Shape-from-Motion to a deformable environment for about a decade, yet with only very limited success to date. We believe that there are two main reasons for this. Firstly there is still a lack of a solid theory for Deformable Shape-from-Motion. Fundamental questions, such as what kinds of deformation can facilitate unambiguous 3D reconstruction, are not yet answered. Secondly practical solutions have not yet come about: for accurate and dense 3D shape results, the Motion cue must be combined with other visual cues, since it is certainly weaker in the deformable case. It may require strong object-specific priors, needing one to bridge the gap with object recognition. This project develops these two key areas. It includes three main lines of research: theory, its computational implementation, and its real-world application. Deformable Multiple-View Geometry will generalize the existing rigid theory and will provide researchers with a rigorous mathematical framework that underpins the use of Motion as a proper visual cue for Deformable 3D Reconstruction. Our theory will require us to introduce new mathematical tools from differentiable projective manifolds. Our implementation will study and develop new computational means for solving the difficult inverse problems formulated in our theory. Finally, we will develop cutting-edge applications of our framework specific to Minimally Invasive Surgery, for which there is a very high need for 3D computer vision.

1 481 294 €

Start date: 2013-01-01, End date: 2018-12-31

Using recent progress in laser technology and in particular in the field of ultra-fast lasers, we are getting close to accomplish the alchemist dream of transforming materials. Compact lasers can generate pulses with ultra-high peak powers in the Tera-Watt or even Peta-Watt ranges. These high-power pulses lead to a radically different laser-matter interaction than the one obtained with conventional lasers. Non-linear multi-photons processes are observed; they open new and exciting opportunities to tailor the matter in its intimate structure with sub-wavelength spatial resolutions and in the three dimensions. This project is aiming at exploring the use of these ultrafast lasers to locally tailor the physical properties of glass materials. More specifically, our objective is to create polymorphs embedded in bulk structures and to demonstrate their use as means to introduce new functionalities in the material. The long-term objective is to develop the scientific understanding and technological know-how to create three-dimensional objects with nanoscale features where optics, fluidics and micromechanical elements as well as active functions are integrated in a single monolithic piece of glass and to do so using a single process. This is a multidisciplinary research that pushes the frontier of our current knowledge of femtosecond laser interaction with glass to demonstrate a novel design platform for future micro-/nano- systems.

1 757 396 €

Start date: 2012-12-01, End date: 2017-11-30

Miniaturized cantilever–like sensors have evolved rapidly. However, when it comes to major breakthroughs in both fundamental studies as well as commercial applications these sensors face severe challenges: i) reliability – often only one or two measurements are performed for the same conditions due to very slow data generation and the results are rarely confirmed by orthogonal sensing technologies, ii) sensitivity – in many applications the need is now for ultra-low sensitivities, iii) reproducibility – very few results have been reported on reproducibility of these sensors iv)throughput –extremely slow and tedious read-out technologies. In order to take a great leap forward in cantilever-like sensing I suggest a new generation of simplified and optimized cantilever-like sensing structures implemented in a DVD based platform which will specifically address these issues. My overall hypothesis is that the true potential of these exciting sensors can only be released when using a simple and reliable read-out system that allows us to focus on the mechanical performance of the sensors. Thus we will keep the sensors as simple as possible. The DVD readout makes it possible to generate large amount of data and to focus on mechanics and the interplay between mechanics, optics and electrochemistry. It will be a technological challenge to realize a robust and reliable DVD platform, that facilitates optical read-out as well as actuation. The DVD platform will enable a fast and iterative development of hybrid cantilever-like systems which draw upon our more than 10 years experience in the field. These sensors will be realised using Si and polymer based cleanroom fabrication. Focus is on design, fabrication, characterization and applications of cantilever-like sensors and on DVD inspired system integration. By the end of HERMES we will have a unique platform which will be the onset of many new types of specific high –throughput applications and sensor development projects.

2 499 466 €

Start date: 2013-02-01, End date: 2018-01-31

"I propose implicit programming, a paradigm for developing reliable software using new programming language specification constructs and tools, supported through the new notion of software synthesis procedures. The paradigm will enable developers to use specifications as executable programming language constructs and will automate some of the program construction tasks to the point where they become feasible for the end users. Implicit programming will increase developer productivity by enabling developers to focus on the desired software functionality instead of worrying about low-level implementation details. Implicit programming will also improve software reliability, because the presence of specifications will make programs easier to analyze. From the algorithmic perspective, I propose a new agenda for research in algorithms for decidable logical theories. An input to such an algorithm is a logical formula (or a boolean-valued programming language expressions). Whereas a decision procedure for satisfiability merely checks whether there exists a satisfying assignment for the formula, we propose to develop synthesis procedures. A synthesis procedure views the input as a relation between inputs and outputs, and produces a function from input variables to output variables. In other words, it transforms a specification into a computable function. We will design synthesis procedures for important classes of formulas motivated by useful programming language fragments. We will use synthesis procedures as a compilation mechanism for declarative programming language constructs, ensuring correctness by construction. To develop practical synthesis procedures we will combine insights from decision procedure research (including the results on SMT solvers), with the research on compiler construction, program analysis, and program transformation. The experience from the rich model toolkit initiative (http://RichModels.org) will help us address these goals."

1 439 240 €

Start date: 2012-12-01, End date: 2017-11-30

Living tissues are regulated by multi-cellular collectives mediated at cellular level through complex interactions between mechanical and biochemical factors. A further understanding of these mechanisms could provide new insights in the development of therapies and diagnosis techniques, reducing animal experiments. I propose a combined and complementary methodology to advance in the knowledge of how cells interact with each other and with the environment to produce the large-scale organization typical of tissues. I will couple in-silico and in-vitro models for investigating the micro-fabrication of tissues in-vitro using a 3D multicellular environment. By computational cell-based modelling of tissue development, I will use a multiscale and multiphysics approach to investigate various key factors: how environmental conditions (mechanical and biochemical) drive cell behaviour, how individual cell behaviour produces multicellular patterns, how cells respond to the multicellular environment, how cells are able to fabricate new tissues and how cell-matrix interactions affect these processes. In-vitro experiments will be developed to validate numerical models, determine their parameters, improve their hypotheses and help designing new experiments. The in-vitro experiments will be performed in a microfluidic platform capable of controlling biochemical and mechanical conditions in a 3D environment. This research will be applied in three applications, where the role of environment conditions is important and the main biological events are cell migration, cell-matrix and cell-cell interactions: bone regeneration, wound healing and angiogenesis.

1 299 083 €

Start date: 2012-11-01, End date: 2018-05-31

Advances in numerical methods and three-dimensional imaging techniques have enabled the quantification of cardiovascular mechanics in subject-specific anatomic and physiologic models. Research efforts have been focused mainly on three areas: pathogenesis of vascular disease, development of medical devices, and virtual surgical planning. However, despite great initial promise, the actual use of patient-specific computer modelling in the clinic has been very limited. Clinical diagnosis still relies entirely on traditional methods based on imaging and invasive measurements and sampling. The same invasive trial-and-error paradigm is often seen in vascular disease research, where animal models are used profusely to quantify simple metrics that could perhaps be evaluated via non-invasive computer modelling techniques. Lastly, medical device manufacturers rely mostly on in-vitro models to investigate the anatomic variations, arterial deformations, and biomechanical forces needed for the design of stents and stent-grafts. In this project, I aim to develop an integrated image-based computer modelling framework for subject-specific cardiovascular simulation with dynamically adapting boundary conditions capable of representing alterations in the physiologic state of the patient. This computer framework will be directly applied in clinical settings to complement and enhance current diagnostic practices, working towards the goal of personalized cardiovascular medicine.

1 491 593 €

Start date: 2012-12-01, End date: 2018-11-30

The objective is to develop image processing algorithms for cinema that allow people watching a movie on a screen to see the same details and colors as people at the shooting location can. It is due to camera and display limitations that the shooting location and the images on the screen are perceived very differently. We want to be able to use common cameras and displays (as opposed to highly expensive hardware systems) and work solely on processing the video so that our perception of the scene and of the images on the screen match, without having to add artifical lights when shooting or to manually correct the colors to adapt to a particular display device. Given that in terms of sensing capabilities cameras are in most regards better than human photoreceptors, the superiority of human vision over camera systems lies in the better processing which is carried out in the retina and visual cortex. Therefore, rather than working on the hardware, improving lenses and sensors, we will instead use, whenever possible, existing knowledge on visual neuroscience and models on visual perception to develop software methods mimicking neural processes in the human visual system, and apply these methods to images captured with a regular camera. From a technological standpoint, reaching our goal will be a remarkable achievement which will impact how movies are made (in less time, with less equipment, with smaller crews, with more artistic freedom) but also which movies are made (since good-visual-quality productions will become more affordable.) We also anticipate a considerable technological impact in the realm of consumer video. From a scientific standpoint, this will imply finding solutions for several challenging open problems in image processing and computer vision, but it also has a strong potential to bring methodological advances to other domains like experimental psychology and visual neuroscience.

1 499 160 €

Start date: 2012-10-01, End date: 2018-03-31

"Wearing sub-optimally fitted lower limb prosthesis cause disorders of the stump that strongly lessens the well-being and the performance of an amputee. As experimental measurements are currently not capable of providing enough insights in the dynamic behaviour of the stump, simulations need to be employed to achieve the necessary knowledge gain to significantly improve the socket design and, hence, to increase the amputee’s well-being and performance. The overall goal of this proposal is to provide the enabling technology in form of novel computational and experimental methodologies to assist the design process of next-generation prosthetic devices. The focus hereby is to gain a better understanding of the dynamics of the musculoskeletal system of a lower extremity amputee, here, the stump of an above-knee amputee. To achieve this, LEAD pursues two aims. The first and main aim focuses on substantially changing existing modelling philosophies and methodologies of forward dynamics approaches such that they are capable of representing muscles, bone, and skin as 3D continuum-mechanical objects. To counteract the increase of computational cost by switching from 1D lumped-parameter models to 3D models, novel, elegant, and efficient algorithms, e.g. nested iteration techniques tuned for efficiency through model-based coupling strategies and optimised solvers, need to be developed. The second aim is to experimentally measure physical quantities that provide the necessary input to drive the forward dynamics model, e.g. EMG, and to provide means of validation, e.g. with respect to pressure measurements, ultrasound recordings, and motion capture. Given the non-existing field of forward dynamics appealing to continuum-mechanical skeletal muscle models, LEAD creates a new field of research."

1 676 760 €

Start date: 2012-11-01, End date: 2017-10-31

Angiogenesis, the formation of new blood vessels from the existing vasculature, is a process that is fundamental to normal tissue growth, wound repair and disease. The control of angiogenesis is of utmost importance for tissue regenerative therapies as well as cancer treatment, however this remains a challenge. The extracellular matrix (ECM) is a one of the key controlling factors of angiogenesis. The mechanisms through which the ECM exerts its influence are poorly understood. MAtrix will create unprecedented opportunities for unraveling the role of the ECM in angiogenesis. It will do so by creating a highly innovative, multiscale in silico model that provides quantitative, subcellular resolution on cell-matrix interaction, which is key to the understanding of cell migration. In this way, MAtrix goes substantially beyond the state of the art in terms of computational models of angiogenesis. It will integrate mechanisms of ECM-mediated cell migration and relate them to intracellular regulatory mechanisms of angiogenesis. Apart from its innovation in terms of computational modelling, MAtrix’ impact is related to its interdisciplinarity, involving computer simulations and in vitro experiments. This will enable to investigate research hypotheses on the role of the ECM in angiogenesis that are generated by the in silico model. State of the art technologies (fluorescence microscopy, cell and ECM mechanics, biomaterials design) will be applied –in conjunction with the in silico model- to quantity cell-ECM mechanical interaction at a subcellular level and the dynamics of cell migration. In vitro experiments will be performed for a broad range of biomaterials and their characteristics. In this way, MAtrix will deliver a proof-of-concept that an in silico model can help in identifying and prioritising biomaterials characteristics, relevant for angiogenesis. MAtrix’ findings can have a major impact on the development of therapies that want to control the angiogenic response.

1 497 400 €

Start date: 2013-04-01, End date: 2018-03-31

Computational modelling is changing the face of science. Many complex systems can be understood as embodied computational systems performing distributed computations on a massive scale. Biology is the discipline where these ideas find their most natural application: cells can be viewed as input/ output devices, with proteins and organelles behaving as finite state machines performing distributed computations inside the cell. This led to the influential framework of cell as computation, and the successful deployment of formal verification and analysis on models of biological systems. This paradigm shift in our understanding of biology has been possible due to the increasingly quantitative experimental techniques being developed in experimental biology. Formal modelling techniques, however, do not have mechanisms to directly include the information obtained from experimental observations in a statistically consistent way. This difficulty in relating the experimental and theoretical developments in biology is a central problem: without incorporating observations, it is extremely difficult to obtain reliable parametrisations of models. More importantly, it is impossible to assess the confidence of model predictions. This means that the central scientific task of falsifying hypotheses cannot be performed in a statistically meaningful way, and that it is very difficult to employ model predictions to rationally plan novel experiments. In this project we will build and develop machine learning tools for continuous time stochastic processes to obtain a principled treatment of the uncertainty at every step of the modelling pipeline. We will use and extend probabilistic programming languages to fully automate the inference tasks, and link to advanced modelling languages to allow formal analysis tools to be deployed in a data modelling framework. We will pursue twoapplications to fundamental problems in systems biology, guaranteeing impact on exciting scientific questions.

1 421 944 €

Start date: 2012-10-01, End date: 2017-09-30

"Quantum information science (QIS) is a young research area at the frontier of both computer science and physics. It studies what happens when we apply the principles of quantum mechanics to problems in computer science and information processing. This has resulted in many unexpected discoveries and opened up new frontiers. Quantum algorithms (such as Shor’s factoring algorithm) can solve computational problems that are intractable for conventional computers. Quantum mechanics also enables quantum cryptography which provides an ultimate degree of security that cannot be achieved by conventional methods. These developments have generated an enormous interest both in building a quantum computer and exploring the mathematical foundations of quantum information. We will study computer science aspects of QIS. Our first goal is to develop new quantum algorithms and, more generally, new algorithmic techniques for developing quantum algorithms. We will explore a variety of new ideas: quantum walks, span programs, learning graphs, linear equation solving, computing by transforming quantum states. Secondly, we will study the limits of quantum computing. We will look at various classes of computational problems and analyze what are the biggest speedups that quantum algorithms can achieve. We will also work on identifying computational problems which are hard even for a quantum computer. Such problems can serve as a basis for cryptography that would be secure against quantum computers. Thirdly, the ideas from quantum information can lead to very surprising connections between different fields. The mathematical methods from quantum information can be applied to solve purely classical (non-quantum) problems in computer science. The ideas from computer science can be used to study the complexity of physical systems in quantum mechanics. We think that both of those directions have the potential for unexpected breakthroughs and we will pursue both of them."

1 360 980 €

Start date: 2013-05-01, End date: 2018-04-30

Nowadays, the treatment of cancer is based on the so-called diagnostic paradigm. We feel that the shift from the traditional diagnostic paradigm to a predictive patient-specific one may lead to more effective therapies. Thus, the objective of this project is to introduce predictive models for cancer growth. These predictive models will take the form of mathematical models developed from first principles and the fundamental features of cancer biology. For these models to be useful in clinical practice, we will need to introduce new numerical algorithms that permit to obtain fast and accurate simulations based on patient-specific data. We propose to develop mathematical models using the framework provided by the mixtures theory and the phase-field method. Our model will account for the growth of the tumor and the vasculature that develops around it, which is essential for the tumor to grow beyond a harmless limited size. We propose to develop new algorithms based on Isogeometric Analysis, which is a recent generalization of Finite Elements with several advantages. The use of Isogeometric Analysis will simplify the interface between medical images and the computational mesh, permitting to generate smooth basis functions necessary to approximate higher-order partial differential equations like those that govern cancer growth. Our modeling and simulation tools will be examined and validated by experimental and clinical observations. To accomplish this, we propose to use anonymized patient-specific data through several patient imaging modalities. Arguably, the successful undertaking of this project, would have the potential to transform classical population/statistics-based treatments of cancer into patient-specific therapies. This would elevate mathematical modeling and simulation of cancer growth to a stage in which it can be used as a quantitatively accurate predictive tool with implications for clinical practice, clinical trial design, and outcome prediction.

1 405 420 €

Start date: 2012-10-01, End date: 2017-09-30

The advent of molecular reprogramming and the associated opportunities for personalised and therapeutic medicine requires the development of novel systems for on-demand delivery of reprogramming factors into cells in order to modulate their activity/identity. Such triggerable systems should allow precise control of the timing, duration, magnitude and spatial release of the reprogramming factors. Furthermore, the system should allow this control even in vivo, using non-invasive means. The present project aims at developing triggerable systems able to release efficiently reprogramming factors on demand. The potential of this technology will be tested in two settings: (i) in the reprogramming of somatic cells in vitro, and (ii) in the improvement of hematopoietic stem cell engraftment in vivo, at the bone marrow. The proposed research involves a team formed by engineers, chemists, biologists and is highly multidisciplinary in nature encompassing elements of engineering, chemistry, system biology, stem cell technology and nanomedicine.

1 699 320 €

Start date: 2012-11-01, End date: 2017-10-31

A paradigm shift is emerging in spacecraft engineering from single, large, and multifunctional satellites towards cooperating groups of small satellites. This will enable innovative applications in areas like Earth observation or telecommunication. Related interdisciplinary research in the field of formation control and networked satellites are key challenges of this proposal. Modern miniaturization techniques allow realization of satellites of continuously smaller masses, thus enabling cost-efficient implementation of distributed multi-satellite systems. In preparation my team has already realized two satellites at only 1 kg mass in the University Würzburg’s Ex¬perimental satellite (UWE) program, emphasizing crucial components for formation flying, like communication (UWE-1, launched 2005), attitude determination (UWE-2, launched 2009), and attitude control (UWE-3, launched 2013). My vision for the proposed project is to demonstrate formation control of four pico-satellites in-orbit for the first time worldwide. To realize this objective, innovative multi-satellite networked orbit control based on relative position and attitude of each satellite is to be implemented in order to enable Earth observations based on multipoint measurements. Related sensor systems used in my laboratory in research for advanced characterization of teams of mobile robots will be transferred to the space environment. Breakthroughs are expected by combining optimal control strategies for coordination of relative motion with a robust flow of information in the network of satellites and ground stations, implemented via innovative use of ad-hoc networks in space. Based on my team’s expertise in implementing very small satellites, first time a system composed of four satellites will be launched to demonstrate autonomous distributed formation control in orbit. This research evaluation in space is expected to open up significant application potential for future distributed satellite system services in Earth observation.

2 500 000 €

Start date: 2014-08-01, End date: 2019-07-31

In machine learning and exploratory data analysis, the major goal is the development of solutions for the automatic and efficient extraction of knowledge from data. This ability is key for further progress in science and engineering. A large class of data analysis methods is based on linear eigenproblems. While linear eigenproblems are well studied, and a large part of numerical linear algebra is dedicated to the efficient calculation of eigenvectors of all kinds of structured matrices, they are limited in their modeling capabilities. Important properties like robustness against outliers and sparsity of the eigenvectors are impossible to realize. In turn, we have shown recently that many problems in data analysis can be naturally formulated as nonlinear eigenproblems. In order to use the rich structure of nonlinear eigenproblems with an ease similar to that of linear eigenproblems, a major goal of this proposal is to develop a general framework for the computation of nonlinear eigenvectors. Furthermore, the great potential of nonlinear eigenproblems will be explored in various application areas. As the scope of nonlinear eigenproblems goes far beyond data analysis, this project will have major impact not only in machine learning and its use in computer vision, bioinformatics, and information retrieval, but also in other areas of the natural sciences.

1 271 992 €

Start date: 2012-10-01, End date: 2017-09-30

"Even after more than one century of flight, both civil and military aircraft are still plagued by major vibration problems. A well-known example is the external-store induced flutter of the F-16 fighter aircraft. Such dynamical phenomena, commonly known as aeroelastic instabilities, result from the transfer of energy from the free stream to the structure and can lead to limit cycle oscillations, a phenomenon with no linear counterpart. Since nonlinear dynamical systems theory is not yet mature, the inherently nonlinear nature of these oscillations renders their mitigation a particularly difficult problem. The only practical solution to date is to limit aircraft flight envelope to regions where these instabilities are not expected to occur, as verified by intensive and expensive flight campaigns. This limitation results in a severe decrease in both aircraft efficiency and performance. At the heart of this project is a fundamental change in paradigm: although nonlinearity is usually seen as an enemy, I propose to control - and even suppress - aeroelastic instability through the intentional use of nonlinearity. This approach has the potential to bring about a major change in aircraft design and will be achieved thanks to the development of the nonlinear tuned vibration absorber, a new, rigorous nonlinear counterpart of the linear tuned vibration absorber. This work represents a number of significant challenges, because the novel functionalities brought by the intentional use of nonlinearity can be accompanied by adverse nonlinear dynamical effects. The successful mitigation of these unwanted nonlinear effects will be a major objective of our proposed research; it will require achieving both theoretical and technical advances to make it possible. A specific effort will be made to demonstrate experimentally the theoretical findings of this research with extensive wind tunnel testing and practical implementation of the nonlinear tuned vibration absorber. Finally, nonlinear instabilities such as limit cycle oscillations can be found in a number of non-aircraft applications including in bridges, automotive disc brakes and machine tools. The nonlinear tuned vibration absorber could also find uses in resolving problems in these applications, thus ensuring the generic character of the project."

1 316 440 €

Start date: 2012-09-01, End date: 2017-08-31

"The main goal of this project is to lay the foundations of a ``non-polynomial time theory of approximation"" -- the Parameterized Approximation for NP-hard optimization problems. A combination that will use the salient features of Approximation Algorithms and Parameterized Complexity. In the former, one relaxes the requirement of finding an optimum solution. In the latter, one relaxes the requirement of finishing in polynomial time by restricting the combinatorial explosion in the running time to a parameter that for reasonable inputs is much smaller than the input size. This project will explore the following fundamental question: Approximation Algorithms + Parameterized Complexity=? New techniques will be developed that will simultaneously utilize the notions of relaxed time complexity and accuracy and thereby make problems for which both these approaches have failed independently, tractable. It is however conceivable that for some problems even this combined approach may not succeed. But in those situations we will glean valuable insight into the reasons for failure. In parallel to algorithmic studies, an intractability theory will be developed which will provide the theoretical framework to specify the extent to which this approach might work. Thus, on one hand the project will give rise to algorithms that will have impact beyond the boundaries of computer science and on the other hand it will lead to a complexity theory that will go beyond the established notions of intractability. Both these aspects of my project are groundbreaking -- the new theory will transcend our current ideas of efficient approximation and thereby raise the state of the art to a new level."

1 690 000 €

Start date: 2013-01-01, End date: 2017-12-31

Passwords, passphrases and PINs have become a usability disaster. Even though they are convenient for implementers, they have been over-exploited, and are now increasingly unmanageable for end users, as well as insecure. The demands placed on users (passwords that are unguessable, all different, regularly changed and never written down) are no longer reasonable now that each person has to manage dozens of passwords. This project will develop and evaluate an alternative design based on a hardware token called Pico that relieves the user from having to remember passwords and PINs. Besides relieving the user from memorization efforts, the Pico solution scales to thousands of credentials, provides ``continuous authentication'' and is resistant to brute force guessing, dictionary attacks, phishing and keylogging. To promote adoption and interoperability, the Pico design has not been patented. The Principal Investigator has been invited to speak about Pico in three continents (including at USENIX Security 2011) since releasing the first draft of his design paper.

1 350 000 €

Start date: 2013-02-01, End date: 2017-12-31

"Quantum Information Processing has the potential to revolutionize the future of information technologies. My long-term vision is a network of quantum and classical devices, where individual agents have the ability to communicate efficiently in a variety of ways with trusted and untrusted parties and securely delegate computational tasks to a number of untrusted large-scale quantum computing servers. In such an interconnected world, the notion of security against malicious adversaries is an imperative. In addition, the interaction between agents must remain efficient and it is important to provide the agents with incentives for honest behaviour. The realization of such a complex network of classical and quantum communication must rely on a solid theoretical foundation that nevertheless is able to foresee and handle the intricacies of real-life implementations. The targeted breakthrough of this proposal is to set the benchmark, both theoretically as well as experimentally, of some necessary communication components for such a hybrid network. The concrete objectives of our project are to: design novel cryptographic primitives as a powerful quantum mechanical toolkit for the quantum network infrastructure; study quantum communication and games in order to minimize the communication overhead and guarantee the honest behaviour of rational agents; enhance the understanding of the fundamentals of quantum mechanics and of classical complexity theory through the lens of quantum information and cryptography; study the security of cryptographic primitives in realistic conditions and use state-of-the-art photonic systems to implement complex cryptographic primitives. An ERC Starting Grant will enable me to reach the above objectives through the creation and coordination of an independent quantum cryptography group that will raise the level of competence and competitiveness of the EU and will provide methods and applications essential for the future of information technology."

980 640 €

Start date: 2013-05-01, End date: 2018-04-30

"It is a matter of strategic independence for Europe to urgently find processes taking into account environmental and economic issues, when mining and recycling rare earths. Currently THERE ARE NO SUCH INDUSTRIAL PROCESS AVAILABLE and 0% WASTE RECYCLING of RARE EARTH ELEMENTS (REE). Plus, 97% of the mining operations are performed in China, hence representing a major Sword of Damoclès for the rest of the world’s economy. We propose to develop a new, cost effective and environmentally friendly REE recycling process. We will achieve this: (i) by enabling, for the first time ever, the fast measurement of free energy of mass transfer between complex fluids; hence it will now be possible to explore an extensive number of process formulations and phase diagrams (such a study usually takes years but will then be performed in a matter of days); (ii) develop predictive models of ion separation including the effect of long-range interactions between metal cations and micelles; (iii) by using the experimental results and prediction tools developed, to design an advanced & environmentally friendly process formulations and pilot plant; (iv) by enhancing the extraction kinetics and selectivity, by implementing a new, innovative and selective triggering cation exchange process step (ca. the exchange kinetics of a cation will be greatly enhance when compared to another one). This will represent a major breakthrough in the field of transfer methods between complex fluids. An expected direct consequence of REE-CYCLE will be that acids’ volumes and other harmful process wastes, will be reduced by one to two orders of magnitude. Furthermore, this new understanding of mechanisms involved in selective ion transfer should open new recycling possibilities and pave the way to economical recovery of metals from a very rapidly growing “mine”, i.e. the diverse metal containing “wastes” generated by used Li-ion batteries, super-capacitors, supported catalysts and fuel cells."

2 255 515 €

Start date: 2013-07-01, End date: 2018-06-30

The most common interpretation of Moore's Law is that the number of components on a chip and accordingly the computer performance doubles every two years. At the end of the 20th century, when clock frequencies stagnated at ~3 GHz, and instruction level parallelism reached the phase of diminishing returns, industry turned towards multiprocessors, and thread level parallelism. However, too much of the technological complexity of multicore architectures is exposed to the programmers, leading to a software development nightmare. We propose a radically new concept of parallel computer architectures, using a higher level of abstraction, Instead of expressing algorithms as a sequence of instruction, we will group instructions into higher-level tasks that will be automatically managed by the architecture, much in the same way superscalar processors managed instruction level parallelism. We envision a holistic approach where the parallel architecture is partially implemented as a software runtime, and the reminder in hardware. The hardware gains the freedom to deliver performance at the expense of additional complexity, as long as it provides the required support primitives for the runtime software to hide complexity from the programmer. Moreover, it offers a single solution that could solve most of the problems we encounter in the current approaches: handling parallelism, the memory wall, the power wall, and the reliability wall in a wide range of application domains from mobile up to supercomputers . We will focus our research on a most efficient form of multicore architecture coupled with vector accelerators for exploiting both thread and data level parallelism. All together, this novel approach toward future parallel architectures is the way to ensure continued performance improvements, getting us out of the technological mess that computers have turned into, once more riding on Moore's Law.

2 356 467 €

Start date: 2013-04-01, End date: 2018-03-31

"We address one of the fundamental challenges of our time: Acting effectively while facing a deluge of data. Massive volumes of data are generated from corporate and public sources every second, in social, scientific and commercial applications. In addition, more and more low level sensor devices are becoming available and accessible, potentially to the benefit of myriads of applications. However, access to the data is limited, due to computational, bandwidth, power and other limitations. Crucially, simply gathering data is not enough: we need to make decisions based on the information we obtain. Thus, one of the key problems is: How can we obtain most decision-relevant information at minimum cost? Most existing techniques are either heuristics with no guarantees, or do not scale to large problems. We recently showed that many information gathering problems satisfy submodularity, an intuitive diminishing returns condition. Its exploitation allowed us to develop algorithms with strong guarantees and empirical performance. However, existing algorithms are limited: they cannot cope with dynamic phenomena that change over time, are inherently centralized and thus do not scale with modern, distributed computing paradigms. Perhaps most crucially, they have been designed with the focus of gathering data, but not for making decisions based on this data. We seek to substantially advance large-scale adaptive decision making under partial observability, by grounding it in the novel computational framework of adaptive submodular optimization. We will develop fundamentally new scalable techniques bridging statistical learning, combinatorial optimization, probabilistic inference and decision theory to overcome the limitations of existing methods. In addition to developing novel theory and algorithms, we will demonstrate the performance of our methods on challenging real world interdisciplinary problems in community sensing, information retrieval and computational sustainability."

1 499 900 €

Start date: 2012-11-01, End date: 2017-10-31

Despite the advances in fabrication technologies such as 3D printing, we still lack the software allowing for anyone to easily manipulate and create useful objects. Not many people possess the required skills and time to create elegant designs that conform to precise technical specifications. 'By--example' shape synthesis methods are promising to address this problem: New shapes are automatically synthesized by assembling parts cutout of examples. The underlying assumption is that if parts are stitched along similar areas, the result will be similar in terms of its low--level representation: Any small spatial neighborhood in the output matches a neighborhood in the input. However, these approaches offer little control over the global organization of the synthesized shapes, which is randomized. The ShapeForge challenge is to automatically produce new objects visually similar to a set of examples, while ensuring that the generated objects can enforce a specific purpose, such as supporting weight distributed in space, affording for seating space or allowing for light to go through. This properties are crucial for someone designing furniture, lamps, containers, stairs and many of the common objects surrounding us. The originality of my approach is to cast a new view on the problem of 'by--example' shape synthesis, formulating it as the joint optimization of 'by--example' objectives, semantic descriptions of the content, as well as structural and fabrication objectives. Throughout the project, we will consider the full creation pipeline, from modeling to the actual fabrication of objects on a 3D printer. We will test our results on printed parts, verifying that they can be fabricated and exhibit the requested structural properties in terms of stability and resistance.

1 301 832 €

Start date: 2012-12-01, End date: 2017-11-30

The development of faster, cheaper and smaller transistors has been the driving force behind the exponential growth in computing power over the past 50 years. While our ability to fabricate better transistors has not yet ceased, continuing to translate these advances into better system-level performance is now a major challenge. This proposal seeks to research a new approach to building programmable digital systems, one that can offer the efficiency, robustness and flexibility required as we approach the end of the CMOS era and start to introduce new post-CMOS technologies. The ideas are centred upon a novel network-centric multiprocessor architecture, with contributions planned at every level from the circuit to the language level.

1 271 216 €

Start date: 2012-09-01, End date: 2018-05-31

"This project leads to the development of novel MEAs comprising components elaborated by the electrospinning technique. Proton exchange membranes will be elaborated from electrospun ionomer fibres and characterised. In the first stages of the work, we will use commercial perfluorosulfonic acid polymers, but later we will extend the study to specific partially fluorinated ionomers developed within th project, as well as to sulfonated polyaromatic ionomers. Fuel cell electrodes will be prepared using conducting fibres prepared by electrospinning as supports. Initially we will focus on carbon nanofibres, and then on modified carbon support materials (heteroatom functionalisation, oriented carbons) and finally on metal oxides and carbides. The resultant nanofibres will serve as support for the deposition of metal catalyst particles (Pt, Pt/Co, Pt/Ru). Conventional impregnation routes and also a novel “one pot” method will be used. Detailed (structural, morphological, electrical, electrochemical) characterisation of the electrodes will be carried out in collaboration between partners. The membranes and electrodes developed will be assembled into MEAs using CCM (catalyst coated membrane) and GDE (gas diffusion electrode) approaches and also an original ""membrane coated GDE"" method based on electrospinning. Finally the obtained MEAs will be characterised in situ in an operating fuel cell fed with hydrogen or methanol and the results compared with those of conventional MEAs."

1 352 774 €

Start date: 2013-01-01, End date: 2018-06-30

Large graphs appear in many application areas. Typical examples are the webgraph, the internet graph, friendship graphs of social networks like facebook or Google+, citation graphs, collaboration graphs, and transportation networks. The structure of these graphs contains valuable information but their size makes them very hard to analyze. We need special algorithm that analyze the graph structure via random sampling. The main objective of the proposed project is to advance our understanding of the foundations of sampling processes for the analysis of the structure of large graphs. We will use the approach of Property Testing, a theoretical framework to analyze such sampling algorithms.

1 475 306 €

Start date: 2012-12-01, End date: 2018-11-30

"Learning how to act optimally in high-dimensional stochastic dynamic environments is a fundamental problem in many areas of engineering and computer science. The basic setup is that of an agent who interacts with an environment trying to maximize some long term payoff while having access to observations of the state of the environment. A standard approach to solving this problem is the Reinforcement Learning (RL) paradigm in which an agent is trying to improve its policy by interacting with the environment or, more generally, by using different sources of information such as traces from an expert and interacting with a simulator. In spite of several success stories of the RL paradigm, a unified methodology for scaling-up RL has not emerged to date. The goal of this research proposal is to create a methodology for learning and acting in high-dimensional stochastic dynamic environments that would scale up to real-world applications well and that will be useful across domains and engineering disciplines. We focus on three key aspects of learning and optimization in high dimensional stochastic dynamic environments that are interrelated and essential to scaling up RL. First, we consider the problem of structure learning. This is the problem of how to identify the key features and underlying structures in the environment that are most useful for optimization and learning. Second, we consider the problem of learning, defining, and optimizing skills. Skills are sub-policies whose goal is more focused than solving the whole optimization problem and can hence be more easily learned and optimized. Third, we consider changing the natural reward of the system to obtain desirable properties of the solution such as robustness, adversity to risk and smoothness of the control policy. In order to validate our approach we study two challenging real-world domains: a jet fighter flight simulator and a smart-grid short term control problem."

1 500 000 €

Start date: 2013-01-01, End date: 2017-12-31

This proposal aims via a comprehensive and interdisciplinary programme of research to determine the full response of monolayer (atomic thickness) graphene to extreme axial tensional deformation up to failure and to measure directly its tensile strength, stiffness, strain-to-failure and, most importantly, the effect of orthogonal buckling to its overall tensile properties. Already our recent results have shown that graphene buckling of any form can be suppressed by embedding the flakes into polymer matrices. We have indeed quantified this effect for any flake geometry and have produced master curves relating geometrical aspects to compression strain-to-failure. In the proposed work, we will make good use of this finding by altering the geometry of the flakes and thus design graphene strips (micro-ribbons) of specific dimensions which when embedded to polymer matrices can be stretched to large deformation and even failure without simultaneous buckling in the other direction. This is indeed the only route possible for the exploitation of the potential of graphene as an efficient reinforcement in composites. Since orthogonal buckling during stretching is expected to alter- among other things- the Dirac spectrum and consequently the electronic properties of graphene, we intend to use the technique of Raman spectroscopy to produce stress/ strain maps in two dimensions in order to quantify fully this effect from the mechanical standpoint. Finally, another option for ironing out the wrinkles is to apply a simultaneous thermal field during tensile loading. This will give rise to a biaxial stretching of graphene which presents another interesting field of study particularly for already envisaged applications of graphene in flexible displays and coatings.

2 025 600 €

Start date: 2013-06-01, End date: 2018-05-31

Traffic congestion on motorways is a serious threat for the economic and social life of modern societies and for the environment, which calls for drastic and radical solutions. Conventional traffic management faces limitations. During the last decade, there has been an enormous effort to develop a variety of Vehicle Automation and Communication Systems (VACS) that are expected to revolutionise the features and capabilities of individual vehicles. VACS are typically developed to benefit the individual vehicle, without a clear view for the implications, advantages and disadvantages they may have for the accordingly modified traffic characteristics. Thus, the introduction of VACS brings along the necessity and growing opportunities for adapted or utterly new traffic management. It is the main objective of TRAMAN21 to develop the foundations and first steps that will pave the way towards a new era of motorway traffic management research and practice, which is indispensable for exploiting the evolving VACS deployment. TRAMAN21 assesses the relevance of VACS for improved traffic flow and develops specific options for a sensible upgrade of the traffic conditions, particularly at the network’s weak points, i.e. at bottlenecks and incident locations. The proposed work comprises the development of new traffic flow modelling and control approaches on the basis of appropriate methods from many-particle Physics, Automatic Control and Optimisation. A field trial is included, aiming at a preliminary testing and demonstration of the developed concepts. TRAMAN21 risk stems from the uncertainty in the VACS evolution, which is a challenge for the required modelling and control developments. But, if successful, TRAMAN21 will contribute to a substantial reduction of the estimated annual European traffic congestion cost of 120 billion € and related environmental pollution and will trigger further innovative developments and a new era of traffic flow modelling and control research.

1 496 880 €

Start date: 2013-03-01, End date: 2018-02-28

Globally 1 in 100 children are born with significant congenital heart defects (CHD), representing either new genetic mutations or epigenetic insults that alter cardiac morphogenesis in utero. Embryonic CV systems dynamically regulate structure and function over very short time periods throughout morphogenesis and that biomechanical loading conditions within the heart and great-vessels alter morphogenesis and gene expression. This proposal has structured around a common goal of developing a comprehensive and predictive understanding of the biomechanics and regulation of great-vessel development and its plasticity in response to clinically relevant epigenetic changes in loading conditions. Biomechanical regulation of vascular morphogenesis, including potential aortic arch (AA) reversibility or plasticity after epigenetic events relevant to human CHD are investigated using multimodal experiments in the chick embryo that investigate normal AA growth and remodeling, microsurgical instrumentation that alter ventricular and vascular blood flow loading during critical periods in AA morphogenesis. WP 1 establishes our novel optimization framework, incorporates basic input/output in vivo data sets, and validates. In WP 2 and 3 the numerical models for perturbed biomechanical environment and incorporate new objective functions that have in vivo structural data inputs and predict changes in structure and function. WP 4 incorporates candidate genes and pathways during normal and experimentally altered AA morphogenesis. This proposal develops and validates the first in vivo morphomechanics-integrated three-dimensional mathematical models of AA growth and remodeling that can predict normal growth patterns and abnormal vascular adaptations common in CHD. Multidisciplinary application of bioengineering principles to CHD is likely to provide novel insights and paradigms towards our long-term goal of optimizing CHD interventions, outcomes, and the potential for preventive strategies.

1 995 140 €

Start date: 2013-01-01, End date: 2019-07-31

The goal of computer vision is to interpret complex visual scenes, by recognizing objects and understanding their spatial arrangement within the scene. Achieving this involves learning categories from annotated training images. In the current paradigm, each category is learned starting from scratch without any previous knowledge. This is in contrast with how humans learn, who accumulate knowledge about visual concepts which they reuse to help learning new concepts. The goal of this project is to develop a new paradigm where computers learn visual concepts on top of what they already know, as opposed to learning every concept from scratch. We propose to progressively learn a vast body of visual knowledge, coined Visual Culture, from a variety of available datasets. We will acquire models of the appearance and shape of categories in general, models of specific categories, and models of their spatial organization into scenes. We will start learning from datasets with high degree of supervision and then gradually move to datasets with lower degrees. At each stage we will employ the current body of knowledge to support learning with less supervision. After acquiring Visual Culture from existing datasets, the machine will be ready to learn further with little or no supervision, for example from the Internet. Visual Culture is related to ideas in other fields, but no similar endeavor was undertaken in Computer Vision yet. This project will make an important step toward mastering the complexity of the visual world, by advancing the state-of-the-art in terms of the number of categories that can be localized, and in the variability covered by each model. Moreover, Visual Culture is more than a mere collection of isolated categories, it is is a web of object, background, and scene models connected by spatial relations and sharing visual properties. This will bring us closer to image understanding, the automatic interpretation of complex novel images.

1 481 516 €

Start date: 2013-01-01, End date: 2017-12-31

One of the first things a programmer must commit to in developing any significant piece of software is the representation of the data. In applications where performance or memory consumption is important, this representation is often quite complex: the data may be indexed in multiple ways and use a variety of concrete, interlinked data structures. The current situation, in which programmers either directly write these data structures themselves or use a standard data structure library, leads to two problems: 1:The particular choice of data representation is based on an expectation of what the most common workloads will be; that is, the programmer has already made cost-benefit trade-offs based on the expected distribution of operations the program will perform on these data structures. 2: It is difficult for the programmer to check or even express the high-level consistency properties of complex structures, especially when these structures are shared. This also makes software verification in existing programming languages very hard. We will investigate specification languages for describing and reasoning program data at a much higher level. The hope is that this can reduce the inherited complexity of reasoning about programs. In tandem, we will check if the high level specifications can be semi-automatically mapped specifications to efficient data representations. A novel aspect of our approach allows the user to define global invariants and a restricted set of high level operations, and only then to synthesize a representation that both adheres to the invariants and is highly specialized to exactly the set of operations the user requires. In contrast, the classical approach in databases is to assume nothing about the queries that must be answered; the representation must support all possible operations.

1 577 200 €

Start date: 2013-04-01, End date: 2018-03-31