Luxembourg National Research Fund

Abstract

Computer vision and natural language processing have witnessed several advances in recent years, with unprecedented performance provided by deep representation learning research. Learned artefact representations thus appear attractive to other fields such as Android malware detection, where deep learning on artefact input alleviates the burden on engineering comprehensively hand-crafted features that will be generalize to different malware variants. In this project, named LAAR, we will investigate a new research direction towards representing Android apps that has the potential to become the next frontier in Android app analysis. In particular, we propose to develop novel ways to represent Android dex bytecode, by leveraging the language-like nature of bytecode. Towards successfully producing relevant representations (also called embeddings) to yield a scalable, robust and accurate representation learning of bytecode. By doing so, LAAR could revolutionize the field of machine learning-based Android app analysis, just like BERT revolutionized the NLP field. Concretely, with the yielded representation, we hope to push research on malware detection and dissection, vulnerability identification, repackaging detection, and many other downstream tasks.

Robotics plays a key role in the EU digital strategy, expanding the competitiveness of the industrial sector and offering new solutions to societal challenges. Robots need to deeply understand their current and future situation, relying on what they perceive. They require complex reasoning, decision-making, and autonomous tasks execution. To that purpose, robotic situational awareness poses a promising challenge due to its complex nature. Novel Situational Graphs (S-Graphs) have been proposed by the applicant’s supervisor, as an excellent approach for robotic situational awareness. They can jointly estimate the state of the robot while simultaneously capturing multiple levels of abstraction of the knowledge of the situation.

The main goal of RoboSAUR is to enrich the S-Graphs with complex geometric, semantic, dynamic, and topological information that provides robots with a more powerful situational awareness usable in real applications. Concretely, we will enhance the S-Graphs, exploiting and extending existing machine learning-based algorithms, by:

(i) Adapting existing tools for real-time semantic segmentation and integrating algorithms to efficiently extract deeper semantic (e.g. color, material) and dynamic information leveraging existing common sense.

(ii) Using existing learnable representations for the geometric information, integrating the measurements with prior knowledge (e.g. concrete shape of a chair knowing how chairs are), used for predicting the geometric appearance of unseen sides.

(iii) Automatically building high-level concepts and relations between them that enlarge the S-Graph. All our work will be validated in real experiments for inspection and surveillance with terrestrial and aerial robots.

We will employ the ready-to-use robotics platforms of the research group of the applicant’s supervisor. We expect that RoboSAUR will push the limits of the state-of-the-art towards more intelligent and autonomous robots by providing them with real-time situational awareness with rich multi-abstraction representations. We foresee the publication of, at least, four research papers in prestigious robotics peer-reviewed international journals and conference proceedings, together with a software implementation of the developed robotic situational awareness system with potential for commercialization.

Symmetric cryptography is an indispensable part of virtually all modern security architectures and protocols. In the past 15 years, cryptographic permutations emerged as a new kind of symmetric primitive that differs from block ciphers in two major aspects, namely (i) they are unkeyed (and, thus, they also do not have a key schedule), and (ii) the inverse permutation is normally not required. Permutation-based cryptography started to attract particular interest when the hash function Keccak was announced to become an international cryptographic standard. Permutations are extremely flexible and versatile primitives, similar to block ciphers, and can be used to construct e.g. hash functions, message authentication codes, pseudo-random bit-sequence generators, stream ciphers, and authenticated encryption algorithms.

Benchmarking results from the currently-ongoing lightweight cryptography standardization project of the National Institute of Standards and Technology (NIST) confirm that permutation-based cryptosystems can reach high performance in hardware and software. Among the 10 final-round candidates of the NIST competition are three that use a permutation as underlying primitive; these are the Addition-Rotation-Xor (ARX) design SPARKLE, and the two AND-Rotation-XOR (AndRX) designs ASCON and Xoodyak. The mission of the proposed research project is to extend the current body of knowledge on the design, analysis, and implementation of permutation-based cryptosystems in three different directions.

First, we aim to develop new cryptanalytic techniques for ARX/AndRX designs in general, and the NIST finalists ASCON, SPARKLE, and Xoodyak in particular, which is important because these designs are relatively new and the robustness of SPARKLE and Xoodyak against differential cryptanalysis has not been widely researched until now. We will focus especially on exploring the cryptanalytic similarities and differences between ARX and AndRX designs and try to answer the question whether existing methods for the cryptanalysis of ARX constructions can be adapted (i.e. optimized) for AndRX cryptosystems. In the second line of research, we aim to design an improved variant of the ASCON permutation that overcomes the two weaknesses of the original design, namely (i) the state size of only 320 bits, which is too small to be competitive with SPARKLE and Xoodyak, and (ii) the rotation distances, which are not efficient on small 8 and 16-bit microcontrollers. The goal is to obtain a cross-platform permutation that can achieve record-setting performance in hardware and software on a wide range of 8/16/32/64-bit processors.

Finally, we will analyze how ARX/AndRX-based permutations can be effectively and efficiently protected against physical attacks, whereby we will pay special attention to white-box techniques. We are particularly interested in the question of how classical countermeasures against (partial) leakage of secrets can be integrated into the white-box model in a secure way. All three topics have the potential to produce results with significant real-world impact since the permutation-based NIST finalists have a serious chance to become the next cryptographic standard and get deployed in billions of devices.

A Cyber-Physical System (CPS) is a collection of various types of interacting components. These components can include sensors, which generate signals, and control components, which can contain software that monitors the output of sensors and takes action when necessary. CPS need to be analysed to ensure that they operate with high reliability. An approach that helps us to attain such reliability is Runtime Verification. This approach often involves monitoring the CPS’s behaviour in order to decide whether it is behaving as desired. Despite the clear usefulness of such an approach, it is becoming clear that engineers need to know more than just whether a CPS behaved as desired; they need to know why. The process of determining why a given CPS behaved in a certain way is often referred to as “diagnostics”. For example, a smartphone might have a proximity sensor, which detects if the user’s face is near the screen during a phone call. In this case we can monitor the smartphone’s sensor and its behaviour, which is to turn off the display to avoid undesired touches to the screen and turn it back on when the users’ face it is not nearby.

The issue at hand is that when performing diagnostics (the screen isn’t turned off when the sensor detects user’s face near the display), instead of getting yes/no, it would be useful to get the reason why. Knowing why the system didn’t behave as desired helps the engineer detect and solve the problem faster and more effectively.

The main contribution of this project is a mathematical framework that provides tools for diagnosing the behaviour of a CPS. Such tools are specific to the types of components found in a CPS and include static and dynamic analysis of source code-based components, and signal analysis of the data generated by sensors. Finally, the tools provided by the mathematical framework allow diagnostics to be performed after the CPS in question has finished executing.

Graph Neural Networks (GNNs) have achieved great success in dealing with the structural data, via the crafted network architecture that is out of ordinary from other deep learning networks. However, similar to traditional deep learning, graph neural networks have poor robustness in the face of specially designed adversarial attacks. Under adversarial attacks, small perturbations can affect graph datasets, which can lead to a severe decline in GNN model performance. Therefore, the research on adversarial graph attacks and how to build robust GNN models have received more attention. Our research conducts the most extensive study on existing GNN test generation techniques for robustness evaluation to explore the effectiveness of current robustness test technologies. Besides, we aim to assess current adversarial training data generation strategies on their ability for model robustness enhancement. As the robustness enhancement approaches require a large amount of adversarial train data, we thus expect to perform significant data selection metrics to reduce the number of training data and optimize the graph adversarial training process.

A digital city’s spatial data infrastructure relies on high-quality three-dimensional (3D) models. They have many uses, including urban planning building information modelling, urban assets management, emergency response, autonomous driving, change detection and environmental studies and state cadastral inspection. Airborne Light Detection and Ranging (LiDAR) and aerial photogrammetry are promising means to provide fast and efficient large-scale 3D remote sensing data over the urban area in the form of point clouds. A sufficiently dense, accurate and complete point cloud is the primary data source that can transfer to a photorealistic 3D city model, known as the city level of details (LOD) model. The newly introduced hybrid mobile sensor (i.e., Leica CityMapper 2) led the airborne mapping sector a step forward by integrating aerial LiDAR and photogrammetry (both nadir and oblique-looking) systems on the same airborne platform. However, point clouds from airborne LiDAR and photogrammetry, i.e. dense image matching (DIM), differ significantly in geometric accuracy, precision (e.g., the presence of noise), density, amount and size of data gaps and available attributes. Integration of airborne LiDAR with photogrammetric point clouds is one of the most challenging aspects of geospatial data processing due to variances between two different types of datasets or sensors. The fusion can only be performed if they are registered precisely in order to eliminate the geometric inconsistency.

This research will develop an efficient and reliable method for precise registration between 3D LiDAR points and the photogrammetric imagery during the DIM workflow in order to generate datasets with superior registration accuracy. The first initiative is to bring the image segmentation, point cloud classification and object detection (façade, roof, window, etc.) methods for hybrid sensors data to the next level. Later the fused data will be used to perform machine learning algorithms for automatic point cloud classification because a complete, dense and classified point cloud will considerably impact the automatic way of detail geometry generation to improve 3D city models and the performance of related applications. The substantial impact of this project on detailed geometry generation is also essential for better strategic decision-making in terms of digitization and sustainability.

Connected and Automated Driving (CAD) desires to provide innovative services that enable a vehicle to be better informed about the surrounding environment and make more accurate and faster decisions in critical situations, thus reducing road casualties. Vehicle-to-Everything (V2X) communication facilitates CAD and allows vehicles to communicate with each other. V2X has evolved rapidly over the past few years, offering promising applications ranging from the enhancement of road safety to comfort. One of those applications is the cooperative perception system (CPS), where vehicles and roadside infrastructures exchange sensory information using the V2X network to maximize the perception horizon. V2X can be performed between participants on the CAD ecosystem for dynamic real-time information exchange. However, communicating perception sensors data requires a significant amount of network bandwidth. With the recent deployment of low latency and high throughput network technologies such as 5G in many metropolitan areas, sensor data sharing between vehicles, infrastructures, and beyond is becoming a realistic option. Hence, sensor data sharing could easily be the next key enabler for the vision to minimize existing driving challenges while increasing the awareness horizon beyond local sensing capacities. However, a CPS can also generate lots of redundant data, and data transmission comes at the cost of processing overhead. Therefore, any novel CPS applications must be tested to assure their reliability. Still, trials on actual vehicles are uncommon due to the complexity, cost, and risks associated with the experiments. Simulation appears to be an excellent alternative to address this issue, and the research community frequently relies on it. In this context, the simulation environments should be as realistic as possible and mimic the challenges of an actual deployment. So far, no existing framework combines synthetic car sensor data with the realistic network simulation needed to evaluate CPS.

This Ph.D. project, “Infrastructure assisted cooperative driving strategy for connected vehicles (ACDC),” will focus on designing such a framework that allows us to validate CPS solutions and determine how sensor data could be efficiently shared between vehicles and infrastructures using V2X facilities through a realistic communication channel. We suppose that such a platform will open many exciting research avenues. Another goal of the project is to explore novel distributed learning techniques to distribute sensor information processing among vehicles, infrastructures, and beyond and find the best trade-off between data processing and transmission policies while ensuring low latency and high reliability.

Nowadays the space sector is no longer an exclusive market for governments and military applications. As the access to space has become cheaper, more private companies have entered the space business, increasing its popularity. There even exist companies that use space data without owning any space assets, thanks to services such as satellite-as-a-service. The increase of satellite data and the potential use of the combination of data from more than one satellite data provider makes necessary the design of a system that facilitates this type of operations. In this proposal, we present a novel solution that facilitates and democratize the access to satellite data and increases the opportunity for research and innovation in fields that may take advantage of Earth Observation data. We propose a satellite data marketplace to bring together satellite data providers and satellite data consumers. The aim of this project is to maximize the profits of the mentioned marketplace and reduce the time in which the client receives the data. Profit maximization is divided into two problems, cost minimization and revenue maximization. The cost minimization problem is modeled as a combinatorial problem and as a game theory scenario. To maximize the revenue we propose a machine learning model to predict the demand and set the price that maximizes the revenues. In order to minimize the time in which data gets to the clients, we propose a machine learning model to organize the data considering its demand, data that is more probable to be requested will be stored in a place in which could be retrieved and delivered to the client in the fastest way. The expected outcomes of this project will contribute to national strategies like the national data exchange platform, space strategy and data economy; also, this project is in line with the National Standardization Strategy 2020-2030 and will contribute to current projects of national, European, and international standards and white papers, in collaboration with ILNAS/ANEC, in the context of data usage and artificial intelligence.

Configurable systems form a vast and heterogeneous class of software systems that can be derived into multiple variants (or configurations). A notorious challenge in configurable system engineering is that the number of system variants increases exponentially with the number of configuration options — named features in the jargon. This exponential growth inevitably makes many automated analysis activities — in particular, during Quality Assurance (QA) — intractable for real-world configurable systems. As a result, sampling approaches have been an important support to configurable system QA tasks. Engineers can analyse a representative sample of variants and infer results for other variants. Because variants are too numerous to be all analysed, sampling offers an adequate compromise between completeness and efficiency. Sample uniformity is key to effectively solve the QA tasks, as a biassed sample would decrease the quality of the inference and reduce the accuracy of the results for out-of-sample variants. As the interest towards Uniform Random Sampling (URS) has been growing, different research communities have developed algorithmic solutions and evaluated those solutions on community-specific datasets. These different focuses translate into discrepancies in evaluation benchmarks and conclusions. A recent study has revealed that existing methods, when applied to configurable systems, are *either* efficient *or* uniform. Despite the fact that research has continued to work on new sampling techniques, efficient and uniform random sampling from feature models remains an open challenge. URS methods typically operate from Boolean formulae are typically expressed in Conjunctive Normal Form (CNF). However, research has shown that expressing formulae in deterministic Decomposable Negation Normal Form (d-DNNF) enables the resolution of intractable logic queries in polynomial time. Preliminary investigations we conducted have revealed that, while sampling from d-DNNF is indeed tractable, the conversion from CNF to d-DNNF creates an important bottleneck. Our main hypothesis is that the conversion from CNF to d-DNNF can be improved and yield compressed d-DNNF formulae that enable efficient URS through state-of-the-art sampling algorithms. Hence, the objective of this project is to make state-of-the-art URS algorithms scale to large configurable systems via effective conversion from CNF to d-DNNF. We divide our endeavour into three sub-objectives. First, we will devise a new metric that can measure the complexity of a formula for d-DNNF based sampling. This complexity characterization will help us understand how to reduce the size of a logic formula at affordable computational costs, by pinpointing specific elements that trigger complexity. Second, we will exploit this information and design a semantic-preserving d-DNNF compression method. The principle of our method is to efficiently compute a sound approximation of the Minimal Independent Support (MIS), driven by our complexity metric, which is close to the MIS (includes few redundant variables) and excludes variables that contribute significantly to the initial formula complexity. Doing so, our compression reduces the formula complexity without introducing the significant overhead of computing the exact MIS. Third, we will introduce formal parameters in our method that enable control over the precision of the approximation and the efficiency of the compression (which, in turn, results in trading off sampling uniformity and efficiency). We will then conduct a large-scale experimental study over a large set of formulae and of parameter settings.

Adaptive cruise control (ACC) is a function of advanced driver-assistance systems that extends the cruise (speed) control function so that the controller can keep a safe distance from the vehicle in front of it. The primary objectives of ACC are to improve driving comfort, reduce driving errors, and boost safety. Model Predictive Control (MPC) is currently the most promising solution for the developing ACC, as it can consider several ACC objectives at the same time and incorporate them into the controller in a systematic manner. Weighting factors for the objectives are used to prioritize one objective over another, where some objectives are conflicting. Tuning the weighting factor is highly dependent on the current driving and traffic situation. Therefore, instead of using one tuning for the weighting factors, this project will consider real-time adaptation of the weighting factors based on the driving situation. In addition, MPC-based ACC predicts the behaviour of the own and in front car over a period of time based on models for the system and the environment. Input data are used for prediction; meanwhile, the more information and understanding of the surrounding is, the better the prediction accuracy and thus the higher the performance of the controller. The second aim of this project is to improve the prediction capabilities of the controller to achieve higher performance. To achieve this goal, we propose in the project to use situation awareness, in which information about surroundings can be obtained from the sensors like cameras, radar and GPS. Then the information will be processed and analysed to extract the useful information that can be fed to the controller. In summary, the research will focus on two parts: developing novel situation awareness approaches that are appropriate for ACC applications, and improving and preparing the MPC-based controller to be compatible with situation awareness.

On February 15th, 2022, the EU Commission put forward a €6 Billion initiative for a secure satellite communication system. The press release sets the expectation of a space infrastructure that provides high-speed internet to EU citizens and contributes to a green and resilient future for Europe and the planet. The initiative also recognizes the risks in the exponential growth in orbiting satellites and seeks to establish a Space Traffic Management infrastructure to ensure “space remains a safe, secure, and sustainable environment.” With this initiative, the EU is communicating a satellite future: a vision of a desirable satellite infrastructure that sets expectations for the future and is used to shape the direction of technological and societal development. Such visions encourage, or discourage, various forms of research and development in service of achieving the portrayed future. There is a growing consensus that the communication of the future shapes science, technology, society, politics, and culture in remarkable, interdependent ways. Historians have shown how visions of the future connect scientists, policymakers, and the public, thus facilitating which future is realized. Furthermore, it is argued that the means by which futures are pursued – be it through policy or research focus – affect the way technologies are developed and utilized. SATFUTURE aims to understand the history, and future, of satellite infrastructures, focusing on satellite futures. It will showcase how the communication of the future shapes the way technologies are developed and utilized by asking: How have expectations shaped the development of past satellite infrastructures? In what ways were these expectations incorporated into the final product? And how have these expectations changed over time?

SATFUTURE will contribute to understanding Europe’s history in space by looking back to two of the European Space Agency’s first satellite infrastructures: Meteosat, for meteorology, and the European Remote Sensing Satellite (ERS), for Earth observation. These two case studies offer a unique opportunity to reflect on where Europe’s expectations of space infrastructures once were, and where they stand today.

With economic crises, market collapses and deindustrialisation in the wake of the oil crisis in 1973 in Western Europe and with the transition from planned to market economy and neoliberal turn in economic policy in Central and Eastern Europe during and after 1989/91, major political, economic and social changes have taken place in the last three decades of the 20th century all around Europe. This project combines both phenomena under a common, transnational concept of transformation and postulates, that similarities in the mechanisms and dynamics of these changes can be observed beyond time periods and geographical spaces.

By means of a diachronic comparison of two case studies of the steel company ARBED in Western European Luxembourg and of its subsidiary company the ”Thüringer Stahlwerke” in former East German Unterwellenborn between the years of 1973 and 2001, the project outlines similarities, differences and transnational ties of the transformation. Through an approach that combines business and labour history, archival and media sources and Oral History interviews are used to analyse and compare both the changes in corporate structure and management strategies and the experiences and everyday working practices of workers in both locations. This provides a new transnational understanding of transformation of both business and labour history. The research results and the generated data such as fragments of media and archival sources and of Oral History are also presented within a multimedia online exhibition using digital tools to comparatively visualise the transformation of ARBED in Luxembourg and in Unterwellenborn.

Autism is a lifelong condition yet research to date has focused almost exclusively on children and young adults, leaving a huge gap in our knowledge on what it means to age with autism. The little research that exists on autistic adults has shown that they are less likely to live independently or be employed than the general population, that they die earlier and are more at risk for suicide. This highlights the importance of finding out what aging well means for older autistic adults and what can be done to achieve it. Research with autistic older adults so far has measured quality of life with tools that were designed for and by the general population, which might not accurately reflect the autistic experience. In addition, participants with learning disabilities have been excluded from most studies, even though they represent around 30% of autistic people. Finally, contextual factors which might influence quality of life have been ignored.

Our work will address these limitations and provide new and urgently-needed knowledge on what it means to “age well” as an autistic adult. We will interview autistic adults (including those with learning disabilities) to understand their perspective on what it means to age successfully. We will then design a questionnaire to explore what it means to age well with autism in a larger sample of autistic older adults and their caregivers. This will allow us to identify factors that impact the life satisfaction of older autistic adults. Finally, we will analyse and compare national policy and the services available for autistic adults in the UK, Spain and Luxembourg and how these may be contributing to quality of life. This knowledge can be used as a basis to adapt and improve the support and services provided to autistic people in later life.

In order to successfully tackle climate change and environmental degradation, societies need to shift towards more sustainable modes of production and consumption, replacing harmful carbon-promoting practices with climate- and environment-friendly ones in all industries. The project investigates how the recent reform of the European Union’s farm subsidies programme, the Common Agricultural Policy, contributes to such a transformation in the agri-food sector. Drawing on relevant theories of contemporary policy research, it: examines power relations, coalitions and conflicts that drive and hinder a transition towards low-carbon agriculture; assesses the political feasibility and durability of the farm policy’s green architecture; explores the transition pathway laid out by the policy reform. Applying a novel analytical approach focusing more exclusively on political factors and realities, the research complements standard policy appraisal methods. Its findings will help develop more politically founded policy recommendations and strategies and improve our understanding of sustainability transitions in agriculture.

The base of common knowledge gain is science. However, it is important to ensure that science is conducted well, and potential influential factors held small. Therefore, this project will focus on quality assurance of research, with three main objectives. First, a general issue of social sciences shall be tackled: often psychologists or sociologists rely on self-reports to research people’s behavior, attitudes, thoughts, and feelings, what is problematic, when people do not answer honestly.

When participants present themselves in a socially desirable fashion instead of reporting their true behavior, attitudes and so forth, they introduce a so-called social desirability bias (SoD). It has been found that how self-reports are assessed (f. ex. Personal, online, on the phone) may influence the magnitude of social desirability. Hence it would be important for researchers to know, which is the least impacted assessment mode. It is hypothesized that the more anonymous a mode, the least impacted it is by social desirability. But to find this out, a novel method is needed that synthesizes previous research in a network of evidence and can give a ranking of best to worst, a so-called Bayesian network meta-analysis (BNMA). BNMAs are well established in medicine, but new to social sciences and urgently need transfer. It could help more researchers in social sciences if they knew how to use it. Therefore, after having applied a BNMA to SoD across survey modes, the second objective of the project is to create tutorials for social science researchers, how to conduct BNMAs. The third objective aims at the quality of the future BNMAs. When introducing a new method, it is important to do that in a good practice way. Many sciences recently faced a crisis in which previous findings were not reproduceable, because research practices were not well documented or erroneous. Making mistakes and facing such crisis could be reduced if researchers worked transparent. But it was found that researchers often do not do that, because they do not know how to.

Therefore, guidelines for BNMAs shall be established that researchers can correctly and transparently register how they plan to conduct their BNMA and what outcome they expect before they conduct it. With this project it is hoped to find out how to get more reliable data in self-report research. Moreover, it shall help other researchers make advancements in their field because they can use BNMAs. Lastly, it shall help the quality of research in this field by giving researchers guidelines how to work transparently. All this together can help society because research outcomes are often indicators for policy making.

There is a rapid growth in population aging and consequently the prevalence of age-related diseases such as Parkinson’s disease (PD). PD is a progressive neurodegenerative disorder including non-motor (e.g., depression, memory problem) and motor symptoms (e.g., walking difficulties). Researchers have found that PD symptoms impact patients’ Quality of Life (QoL) and subjective well-being (SWB). SWB is a measure of an individual’s subjective evaluation of his/her life. Given that severity of PD and its impact on patients’ daily life and SWB vary from person to person, it is necessary to develop personalized interventions to improve patients’ level of SWB. This calls for the assessment of SWB in a personalized way to measure the efficacy of the relevant interventions.

This PhD project aims to develop a brief personalized assessment method (PAM) to measure SWB in PD patients which can be applied by both patients and medical doctors. It will be implemented in a digital format that will facilitate patient-doctor interaction, shared decision-making, and improved patient autonomy. The project will pursue this aim by following three interrelated goals. First, the PD symptoms and their impacts on the level of SWB in PD patients will be identified. Second, the specific subgroups of PD patients based on these symptoms/factors will be determined. Third, -based on work related to goals 1 and 2 – a new and brief digital PAM of SWB will develop. The outcome of this PhD project will have tremendous potential among the healthcare professionals to develop personalized interventions for PD patients and measure the efficacy of these interventions in terms of their impact on the level of patients’ SWB.

Fossils, the petrified remains of living beings, represent the archive of life on Earth. They provide unique windows into organismal evolution and ancient ecosystems. Due to the variable preservation potential of organic tissue, however, only organisms with mechanically and chemically resistant body parts such as mineralized shells or skeletons are likely to fossilize. As a result, the fossil record is incomplete by nature, and most palaeontological studies rely on model groups that are particularly well represented in the fossil record.

One of these model groups are the stalked crinoids, or sea lilies. They represent one of the five living classes of echinoderms, alongside starfish and sea urchins, and can be found in virtually every corner of the world’s oceans. Sea lilies, as the name suggests, often have a flower-like appearance and are an essential component of the suspension-feeding communities on the seabed. They have a long and extensive fossil record: remains of stalked crinoids are known from most Palaeozoic and Mesozoic marine rocks, sometimes in such masses that they are rock forming. Because of their rich fossil record, stalked crinoids have been used in the past to explore large-scale evolutionary processes, e.g. the influence of predation pressure on body shape and/or ecological niche occupation or the expulsion of supposedly archaic animal groups into the deep-sea in the course of Earth’s history. An evolutionary study, however, is only as robust as the phylogeny of the underlying organismal group, i.e. the knowledge of the evolutionary relationships among the representatives of that group. The phylogeny of the stalked crinoids is flawed by characters that look superficially similar but show no evolutionary relationships.

In this PhD project, we aim at combining data extracted from the genome and body shape characters of living sea lilies to construct a robust tree of life of the group. First, we will investigate a specific part of the sea lily genome, the so-called transcriptomes, to extract molecular information on the evolutionary ties between the individual species. Once these relationships are resolved, we will look into the details of the sea lily skeleton, employing a variety of microscopy and X-ray scanning methods to unlock the secrets of their microstructures. Ideally, comparing different species and growth stages will tell us which aspects of the body shape are related to shared ancestry and which are the result of independent ecological adaptation.

The characters that unveil the evolutionary ties are the most valuable because they can, in a next step, serve as a basis to re-assess the fossil record of the sea lilies. Including fossils in a tree of life can be a true game changer because fossil data provides a time scale and allows dating of the divergences between species. Ultimately, if all or at least most of our research endeavours prove successful, we will be able to tell the story of the living sea lilies, from their early beginnings 250 million years ago across several mass extinctions, and along an evolutionary trend that led them into their present-day deep-sea harbour. Once this story is told, we can start to read between the lines and use the sea lily story to understand the big patterns in the evolution of life on Earth.

As the third most common form of cancer, colorectal cancer (CRC) is a major healthcare challenge worldwide. While public colorectal cancer (CRC) screening campaigns have yielded great success by preventing disease development through removal of precancerous lesions, many patients are still left with few effective therapeutic options as they often present with an advanced disease stage at diagnosis. Indeed, over 50% of patients are diagnosed in stages III or IV. Immune checkpoint inhibitors are a class of drugs that has changed the way we treat many cancers, but unfortunately, they show little effectiveness for the majority of CRC patients. Indeed, CRC tumors are often described as immunologically cold, being unable to elicit and sustain effective immune responses.

In the tumor microenvironment (TME), there is a web of complex interactions between the tumor cells, immune cells and the surrounding connective tissue(stroma). Within this stromal tissue, the dominant cell type known as cancer associated fibroblast (CAF), has been emerging as a key mediator of the local immune system and shown to be associated with tumor growth. CAF are therefore an important element influencing patient outcomes and treatment effectiveness. One major way in which cells communicate is via the secretion of signalling molecules which interact with receptors on the surface of other cells, thereby inducing them to respond. Type I interferons (IFN-I) are a family of such signalling molecules that are known to have a multitude of effects on the immune system. In the tumour microenvironment, high levels of IFN-I are known to favour the development of tolerogenic niche, meaning a niche where tumor cells are able to evade destruction by immune cells. In our preliminary data from patient derived CRC tumor tissue, we observed that CAF express significantly more IFN-I receptors, making them more sensitive to IFN-I signalling than normal fibroblasts from outside of the tumor environment. Patients whose tumors had a lot of IFN-I signalling also had worse survival rates, and their tumors were enriched with immunosuppressive and exhausted immune cells, making for an ideal environment for uncontrolled tumor growth. In-vitro experiments also showed that stimulation of CAF with IFN-I induced cells to express more of a transmembrane protein (PD-L1) that is known to suppress the adaptive arm of the immune system when stimulated with type I IFN. Based on the reported effects of IFN-I and on our preliminary data, we would like to highlight IFN-I signalling in CAF as an important unexplored but promising target to help improve immunotherapy treatment response in CRC.

Our project will combine state of the art in-vitro and in-vivo approaches and our existing collection of CRC patient samples to identify the role of IFN-I sensitive CAF in therapy resistance and dissect the mechanisms involved.

Within a steadily aging society, the probability of developing a pathological neurodegenerative disease (NDD) beyond the normal and age-appropriate neuronal decline will dramatically increase in the near future. NDD healthcare management is associated with large socioeconomic as well as personal costs, e.g. reduced quality of life within a continuously increasing part of our society. Effective treatments are urgently needed to warrant high quality living conditions in the elderly by delaying disease onset or by slowing down progression.

So far, a major obstacle in the research on NDDs was the limited availability of patient-derived cellular models adequately reflecting brain microenvironment that could be further used for fundamental genetic studies or function as surrogates for pre-clinical testing of potential treatment options. In view of this, the proposed project will utilize the recently developed three-dimensional (3D) brain-organoid model, so-called mini-brains, to capture the multicellular composition of the human brain. The application of oxidative stress as neurotoxic stressor on mini-brains will be used among others to mimic pathological neurodegeneration. To identify the genetic source code essential for the survival of neurons the new CRISPR-Cas9 activation (CRISPRa) technology will be used to identify genes whose activation prevents neuron death under neurotoxic stress and who can be identified within the surviving cell population. No-Piano further aims to determine neuroprotective compounds that hold the potential for direct clinical application by preventing neuronal loss during degeneration. Utilizing the knowledge from the genetic screen, a brain-permeable compound library composed of agents targeting the identified genes associated pathways together with neurotoxic stress will be applied to mini-brains and hopefully lead to the identification of agents conveying neuroprotective properties. Mini-brains originating from NDD-patient cells can further be used in combination with more disease-specific neurotoxic triggers (e.g. Amyloid-Beta, alpha-Synuclein, Tau) to ultimately validate candidate neuroprotective compounds in a NDD-specific context.

Synoptically this project will (1) identify gene-sets and associated pathways exerting a neuroprotective effect in age-related ND-diseases, which could be starting-points for future studies; (2) identify neuroprotective CNS penetrant compounds effective in age-related ND-diseases, which could be further validated in vivo; (3) contribute to translation of fundamental research into applications that can be used for NDD patient care.

Neurodevelopmental disorders (NDDs) which include learning disabilities, autism, and schizophrenia are a major cause of lifelong impairment. Adolescents and adults with NDDs are at a high risk of experiencing symptoms, such as depression, anxiety, memory loss, and hallucinations. However, the biological mechanisms that cause this vulnerability have not been studied. Microglia, the brain’s immune cells, potentially play a key role in the progression of NDDs. We hypothesize that a dysfunction in microglia contributes to the development of symptoms in NDD patients. With advanced PET and MRI scans of the brain, we aim to investigate the role of brain inflammation in the progression of NDDs and the development of symptoms. We will use two mouse models that mimic NDDs and treat them with inflammation stimulating compounds to induce additional worsening of the condition and inflammation suppressing drugs that may prevent disease progression. Behavioral studies will be performed to assess behavioral features of NDDs, which include difficulties with memory, learning, social interactions, and increased anxiety. Changes in brain inflammation detected with PET and MRI scans in the mouse models will form the basis for future PET/MRI studies in NDD patients and help the further investigation of disease mechanisms and treatment possibilities.

All the cells in an organism share the same DNA which stores the information necessary for the cell to function. For the cell to perform the necessary biological processes, the information stored within genes in the DNA is copied into messenger RNA (mRNA) molecules by a process called transcription. Then, each mRNA molecule leads to the production of a protein, by a process called translation. Some proteins can then bind to the DNA in specific sites and activate or inhibit the production of other mRNA molecules in a process called gene regulation. Inside the cell, there are proteins that regulate many genes, and there are genes that are regulated by many proteins. The set of all regulatory interactions is called gene regulatory network (GRN). Understanding how genes interact with each other, referred to as inferring GRNs, is crucial to understand biological systems and how diseases affect the functioning of cells. There are different approaches that try to determine the regulatory interactions of a biological system, from computational predictions to detailed experimental protocols, but the problem still remains challenging. This new era of DNA and RNA sequencing technologies has enabled us to study genes in ways never experienced before, allowing us to have detailed measurements of the more than 20.000 human genes at the same time, for each separate cell of a sample. Despite advent of techniques capable of collecting all this information, GRN inference for all the genes of the genome continues to be extremely complex, due to the high computational costs required by the task of combining all this information. In this project we will develop a combination of mathematical methods that can identify the GRNs efficiently from sequencing data. Specifically, we will use time-series data, which means a set of measurements performed over time, that allows us to understand the chronological response of genes. With the developed pipeline, we will be able to study in depth different diseases, such as Parkinson’s disease, and identify the molecular mechanisms that produce the observable symptoms. As a second case study, we will apply this methodology to study the immune system of crop plants, aiming to understand why different cells have different responses to pathogens. Finally, our method will be applicable to any large-scale dataset. Hence, it will have the potential to help to further understand how complex diseases develop.

Tissue deterioration and death commonly occur during long-term hospitalisation. However, some patients develop tissue injury within less than a week. The onset of such a phenomenon has been associated with severe tissue loading (either in terms of amplitude or duration). Bedsores, ulcers and oedema are similar to bruises on an apple. Oxygen deprivation is also an important factor for the onset of tissue death. Deprived of oxygen, the tissue structure suffers and the tissue “suffocates to death.” Generally, mathematical models do not account for the interplay between both of the above two phenomena, whilst they are known to interact with each other. The objective of our work is to understand the most important mechanical (loading) and bio-chemical (oxygen deprivation) factors responsible for loss of tissue integrity and to couple those within a single model. A unique aspect of our study is to consider tissue death at several spatial and temporal scales. Bio-chemical processes generally occur at the microscopic scale, whereas the mechanical loading is active at the macroscopic scale. As in the case of the bruised apple, the consequences of external loading is visible with the naked eye: interaction of a tissue with an external device (wheel-chair, mattress, prosthesis).

We use the following methods: 1) experimental observation, characterisation (from existing experiments); 2) Modelling and evaluation. Hence, an experimental campaign was conducted on mice, where soft tissue injury was introduced (as in the case of the apple). Experimental results will allow both the identification of parameters of the model to describe properly the tissue, and the evaluation of the model to assess model reliability.

This study will be the basis for future work on animals before moving to organs on chip and humans. It will allow us to make one step toward the next generation patient-specific experimental-mathematical-computational pipeline. This pipeline could become a backbone for a deeper understanding of tissue biophysics and mechanics, enable diagnosis and predictions. Societally, our work has the potential to decrease the burden of ulcers, bedsores and tissue damage in general on patients (wheelchair, mattresses, prosthesis, etc.). Scientifically, we will deliver open access software, data sets and protocols to accelerate future work in computational biomechanics.

The fraction of world population living in water-scarce areas is increasing rapidly, due to the vast rate of the global population growth, the limitation of freshwater resources and additionally because of global warming effects. Beyond that the agriculture, irrigation demand and the water consumption for industrial process additionally stresses the limited fresh water resources. Desalinating seawater represents already today a significant contribution to the provision of additional fresh water sources. In many countries, desalination is the primary source of fresh water. The Middle East and North Africa region houses over half of the world’s desalination capacity. The two major families of the desalination process can be categorized in either thermal or membrane driven processes. Mature technologies as e.g Multi Stage Flash Desalination Systems or Reverse Osmosis System consume a lot of energy and require a sophisticated infrastructure. Approximately half of the world’s desalinated water is generated by membrane technologies. They include techniques like Reverse Osmosis (RO), Membrane distillation (MD), microfiltration (MF), ultrafiltration (UF), and nanofiltration (NF). The key advantages of MD processes over the other conventional technologies are: lower energy costs, lower operating pressure and lower operating temperature. MD has two common configurations, Direct contact (DCMD) and air-gap (AGMD) membrane distillation. The air gap in AGMD provides an additional insulating layer offering advantages regarding the module energy management. On the downside the air gap represents an additional mass transport resistance reducing the fresh water output. In general, the MD technology suffers from lower recovery ratios compared to compete with other desalination technologies.

In recent years a lot of research effort has been invested to increase the MD recovery ratio. Important parameter that influence the permeate flux are the temperature and concentration profile at the interface between the hot feed water channel and the membrane measured by the concentration and temperature polarization of the module. MD module make use of spacers as support structure in the hot feed water channel. These obstacles provide a resistance to the flow and result in a pressure drop in the channel. Similar configurations can be found in other fluid dynamic applications. Investigations of different spacer designs have unrevealed that the pressure drop can be tuned and in addition the flow can be redirected to create eddies impacting the flow boundary size. This can be used to optimize the concentration and temperature boundary layer in order to maximize the permeate flux via the membrane. An additional parameter impacting the bulk temperature as well as the boundary temperature profile is the channel height.

In the proposed study several spacer designs as well as the impact of the channel height will be studied numerically using the ANSYS Fluent CFD solver. The modeling approach used in the work will be validated by experimental boundary layer measurement in the sub- millimeter range. The MeDuSa Test facility at the University of Luxembourg is already equipped with an optical measurement system to determine the temperature boundary. In the course of the project a second light source will be included to measure the temperature and concentration boundary layer independently. The project will produce a unique data set as well as a design modeling method to optimize the hot channel flow to generate a maximized fresh water output.

Contests where prize allocation is delegated to a committee of experts are widespread. In corporate governance, a firm’s owner delegates a CEO hiring decision to a Board of directors. When a bill is introduced, political groups with conflicting interests toward this bill start a competition persuading parliamentarians to vote for or against this bill. In dance sport and figure skating, athletes are ranked based on their performance by a panel of adjudicators. This delegation exists because the contest organizer is not aware about how to evaluate the performance of contestants, the problem she faces is favoritism, when experts can be biased towards a particular contestant due to either intrinsic tastes or affiliations. My research the optimal design of such a contest if the contest organizer simultaneously strives to reward the strongest contestant and maximize the level of competition. Perhaps surprisingly, the rules now commonly implemented in real practice are suboptimal or optimal only under certain conditions, and this project will provide an improvement. This project has a wide applicability in regulation, corporate governance, project and panel evaluations.

Supply chain management is handling various activities that transform raw material into finished products delivered to the end consumer. The activities can range from planning, controlling, and execution. Supply chain activities are generally modelled towards involved parties’ economic performance. However, criteria other than economic performance, such as social and environmental impact, have become important in recent years, especially carbon emission. Because of these, the EU has introduced numerous legislations to reduce carbon emission and increase recycling and reuse of waste streams through the supply chain. It is well known that firms adopting close-loop supply chain (CLSC) including remanufacturing, recycling, and reuse activities are profitable.

In this research, we are concerned with two legislations – i) The European Trading Scheme (ETS) which dictates a 40% reduction of carbon emission by 2030 ii) Extended Producer Responsibilities (EPR) which makes producers physically and financially responsible for end-of-life (EOL) treatment of their product. EPRs that are currently ratified are DIRECTIVE 2019/904 on single-use plastic and Directive 2012/19/EU on waste electrical and electronic equipment (WEEE).

We aim to study the impact of these legislations on Closed Loop SC operations of firms. We consider a setting in which a firm is involved in remanufacturing operations where returned products are remanufactured or dismantled for spare parts. To maximize profit, the firm must acquire (how much and when) and dispose (remanufacturer or dismantle) of the right quantity considering variable return and demand. This problem is much more complex under Carbon emission and EPR legislation. The dilemma is that a higher collection rate (EPR target) pushes for more carbon emissions resulting from reverse logistics (e.g., transportation) and remanufacturing activities.

Finally, we seek to validate our model using a case study on the recycling of aluminum scrap originating from the building construction sector in Luxembourg. Conclusively, the managerial insights from this study will apply to a wide range of products such as Single-use plastics, Li-Batteries, Electronics that are subjected to restrictive policies, particularly in the EU. Conclusively, our research contributes to Luxembourg’s National Plan for Sustainable Development which prioritizes sustainable consumption and production.

The concept of circular economy (CE) has been promoted worldwide to reduce natural resource consumption and waste production in favour of environmental, economic, and social sustainability. The application of CE concepts in urban water management can minimize surface water pollution and groundwater depletion, which is one of the main world environmental concerns due to the population growth and linear pattern of development “take-make-use-dispose”. Personal Care Products (PCPs) have been pointed as emerging contaminants in surface waters and aquatic environment, due mainly to their insufficient removal from discharged Waste Water Treatment Plants (WWTPs) effluents, that thus would require additional and expensive treatment. Scientific studies have been shown that decentralized treatment could be an alternative to reduce their concentration in water bodies. The household’s on-site treatment, for instance, could be also an alternative to remove PCPs from greywater (also called as source control). Besides that, the treated greywater could be reused in sustainable building for toilet flushing and greenery irrigation, reducing the potable-water consumption.

Sustainable solutions as nature-based constructed wetlands (CWs) have been demonstrated a promising option for macro – and micropollutant removal not only from effluent wastewater in WWTP, but also from on-site greywater treatment, by using substrates, and phyto- and bioremediation.

In that context, the main idea of this AFR-PhD proposal is to study the personal care products removal from household’s greywater by using nature-based constructed wetlands with activated biochar as substrate, with focus on specialized microbial communities that could enhance the removal efficiency of these pollutants.

Climate change and environmental degradation are widely considered to be the most concerning global threats of our time, both in terms of potential mass suffering and severe economic dislocation. Given the intensity and scale of climate and nature-related problems at all levels of human life — environmental, economic, political, cultural and social — urgent responses from a range of policy fronts have been called upon. In this context, discussions have been held about the role of central banks in strengthening collective efforts to achieve the 2015 Paris Agreement goals.

The topic has become a priority in international fora but remains surrounded by many uncertainties regarding the extent to which central banks should promote the “greening” of the economy. In light of the prominent position of central banks and their ability to influence economic activity, some academics, international organisations and pressure groups claim that central banks should adjust their operations towards nurturing green investments and limiting activities considered environmentally harmful. On the central banks’ side, studies show that physical risks (climate and nature-related disasters) and transition risks (e.g. the penalisation of carbon-intensive sectors and changes in consumer preferences) pose challenges to the conduct of monetary policy. In this context, a number of studies have shown that monetary policy can indeed reduce climate-induced financial instability and reduce global warming, namely via the deployment of nonstandard tools such as green quantitative easing and collateral allocation.

Notwithstanding the economic and political pressure to deal with the environmental crisis and the seeming existence of tools to achieve that goal, several scholars argue that central banks cannot or should not lead or be responsible for making policy on the climate change. Some of the dangers resulting from assigning central banks a proactive role in this matter relate to the fact that a) a broader mission without a formal and specific mandate can compromise central banks independence, b) a potential stretching of mandates might overburden central banks with conflicting objectives and tools, the prioritisation of which lies in the remit of elected policy makers, c) central banks might be requested to intervene in other pressing areas should the green turn succeed. Regardless the scope of the mandates and/or their interpretation, the evidence shows that an increased number of central banks are rolling out specific measures and proposals to include green aspects into their monetary policy frameworks. In parallel, various green finance initiatives have been set up to enhance the discussion on how to integrate climate risks into mandates and policy processes.

Against this background, the central research question of the proposed research project is: under what institutional conditions can the diverse development of central bank green monetary policies be best understood? Based on a detailed analysis of the existing secondary literature, central bank mandates, news reports, publications and webpages from relevant national and international institutions, the proposed study will perform a comparative political science analysis of the green monetary policy strategies in the ten largest advanced economies and in the ten largest emerging market economies. The subsequent aim will be to undertake a general examination of the green monetary policy measures and proposals up to 2024, contribute to the recent literature on the monetary policy approaches to environmental objectives — in a context where there appears to be inadequate legal basis to legitimise the active role of a number of central banks in greening the economy — and enhance the increased debate on the role of central bank mandates in light of topics that fall outside their conventional goals.

N/A

Elliptic curves are a key research topic in number theory, and they are also currently exploited by cryptography. They are part of algebraic geometry as well: they are curves (geometry) described by a polynomial equation (algebra). Moreover, there is an algebraic operation involving their points and a very rich mathematical structure which led to a vast theory. Elliptic curves are still under investigation, and many interesting questions are open. We plan to focus on elliptic curves defined over number fields or over finite fields, which are connected because one elliptic curve over a number field generates, by the so-called reduction, infinitely many curves over a finite field. The local-global problems see the elliptic curve over the number field as the global object and the other curves as local. Going from local to global is easy, it suffices to accomplish the reduction. However, going from local to global is more mysterious. Properties that hold for all local curves may fail to be true for the global curve (see for example the Conjecture of Linear Dependence, disproved by the supervisor Perucca). Such strange behaviors shed light on the arithmetic of elliptic curves, and we are far from reaching the end of the story. One important point is that elliptic curves are in fashion today, which means that there are plenty of number theorists with which one can discuss mathematics. This is a very flourishing area making great progress every year, and we’d like to be part of it, by joining the research units of Luxembourg and Leiden (a world-class university for this topic, thanks to the mathematical school around Lenstra).