Luxembourg National Research Fund

Abstract

zeroK NanoTech, a US-based scientific instrument company, focuses on developing and commercializing the so-called Low Temperature Ion Source (LoTIS) technology for use in Focused Ion Beam (FIB) applications. With its remarkable performance, the LoTIS source is a key enabler for high lateral resolution SIMS at optimized sensitivity. zeroK and LIST partner in this project to develop a next-generation SIMS instrument, called SIMS:ZERO, a highest-sensitivity, high-speed, ultimate 5 nm resolution SIMS instrument to fulfil the needs of end-users in elemental analysis. zeroK’s LoTIS technology will be used for the primary ion beam while LIST will develop a high-performance SIMS spectrometer for the instrument. The SIMS:ZERO instrument will have a significant national and international impact in the field of nano-analytics and will become a game changer in a number of fields of application.This development will not only help zeroK rapidly capture the ultimate performance SIMS market with its LoTIS technology, but also significantly contribute to the progress of LIST’s activities in high-sensitivity high-resolution nano-analytical instrumentation.

Abstract

The requirements of today’s fiercely competitive markets require better products at shorter innovation cycles. Therefore, a truly new and disruptive virtual design paradigm is urgently needed for which the visionary objective is to advance high performance computing technology for smart virtual prototyping to investigate co-firing of critical fuels in an existing biomass furnace. The XDEM simulation environment allows creating a multi-physics digital twin of a biomass furnace representing the thermal conversion process of both the moving pellets bed and the freeboard. In order to transfer these computationally intensive simulations into an environment for smart virtual prototyping with short production cycles, an innovative so-called overlapping partitioning for arbitrary multi-physics simulation domains is required. It decomposes the simulation domains of the moving fuel bed and the freeboard into overlapping partitions that combines a workload balance with minimal communication overhead between them. Hence, expensive and time-consuming physical prototyping is avoided and a large number of different parameters is easily investigated. In addition, the effect of isolated single parameters is now traceable while other parameters remain constant (ceteris paribus) which physical prototypes simply do not allow. A thorough analysis of predicted results under various co-firing scenarios uncovers casual and hidden relationships that is otherwise not possible in an experimental framework. This newly gained knowledge serves as a basis for developing cutting-edge co-firing technology, is inexpensive and shortens the innovation cycles, which altogether provides a strong advantage in a completive market.

Abstract

Space debris is caused by millions of non-functional, human-made objects left in space that become a hazard for current and future space missions. Yet, no Active Space Debris Removal (ASDR) systems currently exist. Many methods have been proposed, but most of them are still in the technology development stages. Thus, on-ground experimental facilities for the test, verification and validation of ASDR will be critical for pushing the technology to the operational stage. SpaceR and Spacety target to explore, within the HELEN project, the potential of the 2D micro-gravity facility (Zero-G lab) for validating FlexeS, a small Flexible Capture System for debris removal. Advanced computational methods will be developed combining HIL and SIL to recreate high-fidelity in orbit scenarios. The integration of virtual and physical systems will enable close-to-real testing, speeding up the transition between the development and deployment stages of ASDR systems.

Abstract

Our needs in energy are extremely high (160.000 TWh per year) and keep increasing (+ 2 % per year). We have to come up with new strategies and technologies to live in a more sustainable way. In CECOHA, we want to investigate a new energy technology. We will make at LIST active heat exchangers based on Multi Layer Ceramic Capacitors (MLCCs) fabricated by Murata. Heat exchangers are at the heart of thermodynamic machines such as fridges, air conditioning devices and heat energy harvesters. Hence, LIST and Murata recently showed that a heat exchanger based on special MLCCs can induce a temperature gradient of 13 K, a world first (Torello et al., Science, 370, 125 (2020)). This great collective achievement is the starting of this project in which we intend to fabricate a heat pump with a temperature gradient of 20 K and a 1 cm3-heat energy harvester able to generate 1 W of electric energy, both based on bespoke MLCCs from Murata.

ICT: 4 projects 

Abstract

Electronically Assisted Astronomy (EAA) is widely applied today by astronomers to observe planets and faint sky objects like nebulae, galaxies and stars clusters. By capturing raw images from a camera attached to a telescope and processing them on a computing device, this approach allows to generate enhanced views of observed targets that can be displayed in near real time. EAA is also ideal for observers with poor visual acuity, for viewing sessions with family & friends or for public outreach events. In this domain, VAONIS is a fast-growing company specialising in the research and development of a new generation of intelligent telescopes usable by the general public: Stellina and Vespera are fully automated observation stations connected to smartphones or tablets. During the MILAN project (MachIne Learning for AstroNomy), LIST and VAONIS will explore together how recent Deep Learning approaches can help to produce noise-free and realistic astronomical images while going beyond the current hardware limitations and dealing with non-ideal outdoor conditions. Furthermore, the project will contribute to define and apply good practices for embedding Deep Learning models into battery-powered devices with limited computing resources like Raspberry Pi, smartphones or tablets.

Abstract

In banking institutions, text data exchanges in natural language are abundant. To reduce cost and delay, automatically processing textual information is key. Such automation could find applications in various domains such as KYC, email classification, Chatbot creation, etc. Recently, BERT, a new generic language representation model that could be further customized to target a wide range of Natural Language Processing (NLP) tasks, has revolutionized the NLP world. However, Luxembourg specificities call for more! Indeed, language models have shown great promises on “well-formed” English texts. In Luxembourg, the local context (e.g., several languages are mixed) creates challenges decreasing the performance of usual BERT-based models. We propose a collaboration between practitioners and academics. The outcomes will enable our industry partner to better address the request of its clients, e.g., using chatbots or intent analysis in emails. More generally, the development of multilingual models applied to the banking sector will benefit the entire sector and, more broadly, the Luxembourg society.

Abstract

What is true? what is fake? With the increased progress in generating very realistic falsified videos – DeepFakes –, we can no longer rely on videos as evidence. DeepFakes have an immense impact on our society and economy. The world is in a race to find countermeasures, both technological and legal. Currently there is no general solution for detecting Deepfakes. Proposed approaches are specific to a category of artifacts or to a generative adversarial network (GAN). The FakeDeTeR project will exploit spatio-temporal-spectral representations deep learned from facial multimedia data and use them in a classification framework to distinguish fake from true data. Dynamic models for facial data will be used as priors for an effective detection. A higher level of analysis will include sound to video, and exploit cross-modal representations to further strengthen the proposed solution. FakeDeTeR will have a direct impact on cybersecurity and the related industries, by contributing to digital asset authentication and to battling fake news. Indeed, the World Economic Forum ranks the spread of fake news among the world’s top global risks, with an economic cost of $78 billion in 2019.

Abstract

Cloud computing has enabled elastic provisioning of computer resources and revolutionised the industry leading to basically zero cost investment needed for infrastructure of new IT-companies. Most resources provisioned this way come from data centres with a vast array of quite similar compute offerings. With modern distributed client-facing applications, however, there is a need to move the resources closer to the clients, to the Edge. While possible today, it is a manual time-consuming and error prone process. Developers of distributed applications for a modern cloud infrastructure faces two relatively new challenges: the cloud infrastructure is becoming fragmented and heterogeneous with many different kinds of architectures and compute accelerators; it is also becoming much more difficult to deploy and efficiently run these applications as the developers need to know exactly what the infrastructure looks like, both when writing the applications and when they are deployed. The ACE5G project will tackle these challenges by using an execution model which makes different hardware capabilities available at a higher abstraction level, common for all architectures. We will also develop new methods to deploy applications without the need for developers to specify explicitly on which hardware the different components should execute on. These methods will have functionality to also dynamically adapt to changes in requirements and infrastructure and to automatically reschedule parts of application to fulfil requirements while saving resources. We will demonstrate the effectiveness of our approach with several use-cases including the deployment of a software-defined 5G telecom network functionality.

Abstract

Immune checkpoint blockade (ICB)-based cancer immunotherapy is a game changer in terms of overall survival and the cancer patient’s life quality. Despite the exciting and encouraging clinical responses in few patients, majority have a short-term or no survival benefit with severe side effects. The future of immuno-oncology drug development is positioned in combination immunotherapies involving standard cancer therapies rationally combined with ICB’s to boost their efficacy and reduce resistance. The aim of TRICK-ALDH project is to understand the mechanism of action of a groundbreaking combination of ICB’s and specific potent ALDH inhibitors (involved in drug resistance in multiple cancers). This innovation-driven collaborative project between LIH and Advanced BioDesign Pharma Company will undoubtedly contribute to improve the survival benefit of ICB-based cancer immunotherapy in non-responder patients.

Abstract

To help with predicting and mitigating the impacts of droughts, the GRASS project aims to jointly assimilate soil moisture retrieved from Sentinel-1 Synthetic Aperture Radar data and total water storage changes obtained from GRACE&GRACE-FO into a hydrological model. Ultimately, we plan to develop an agricultural drought monitoring service based on agile drone-based remote sensing over Luxembourg. Via advanced hydrological modelling and data assimilation approaches combined with very high-resolution drone monitoring capability, we target short- to medium range predictions of high-resolution soil moisture, groundwater and total water storage, as well as indicators of plant water stress over Luxembourg and the Greater Region. The proposed framework will contribute to: 1) better assess soil water availability; 2) improve drought hazard assessment through the identification of adequate hydrological indices; 3) support evidence-based water resources management; 4) better anticipate and characterise, at a local scale, agricultural droughts that may develop from hydrological droughts. Regarding this last element, the objective is to develop a service leveraging drone-based remote sensing (Visible and Near Infra-Red) campaigns when a hydrological drought is developing. Evidence on an unfolding hydrological drought may be derived from satellite data and hydrological model simulations. Next, the acquired very high-resolution remote sensing data informs on how the event may gradually evolve to an agricultural drought, alongside its potential effects on vegetation. Characterising droughts from a hydrological and an agricultural point of view will be pivotal for stakeholders like farmers, administrations and the risk financing industry – delivering critical information for anticipating drought severity and mitigating its impact. Through the proposed project, we expect to contribute to the development of a (currently missing) operational drought awareness system for Luxembourg.

Abstract

Digital transformation and the COVID-19 economic crisis massively affect low-skilled workers’ labour market opportunities. With LOWSKIM, Jobook, in partnership with LISER, will set up innovative solutions and tailored advice services to recruiters and low-skilled candidates. While most existing recruitment app rely on résumé and focus on high-skilled, this project aims at offering low-skilled workers the access to new job opportunities with a good skill match and with a low risk of automation. LOWSKIM will develop AI solutions in semantic web and clustering technics and conduct field-experiments and surveys to improve the job matching process of Jobook. By exploiting the skill complementarity propensities between occupations across space and business sectors and beware of jobs at risk and growing ones, LOWSKIM will guide job candidates to adequate and secure jobs. LOWSKIM will deliver these innovative and updated services for Luxembourg before jumping to other European countries.

Abstract

Cargolux and the Luxembourg Centre for Logistics and Supply Chain Management at the University of Luxembourg have combined their air cargo and research expertise in CrAFT. The goal of the project is to develop an accurate air cargo turnaround time forecasting tool for cargo airlines to better plan cargo ground operations before an aircraft lands. Better planning increases process reliability, efficiency and sustainability. By leveraging the recent process mapping and extensive data on past turnarounds collected by Cargolux, CrAFT predicts expected turnaround process times for cargo aircraft. We combine econometric forecasting and machine learning prediction, which ensures accurate forecasts that are transparent to users, increasing user trust. Ultimately, Cargolux will implement the developed forecasting model in its IT systems and test it in its operations.