Directed graph (a.k.a. digraph) theory is a lot less developed than (undirected) graph theory and there is a lack of algorithmically meaningful structural theory for digraphs. The objectives of the project is to make some advances on digraph theory in order to get a better understanding of important aspects of digraphs and to have more insight on the differences and the similarities between graphs and digraphs. Our methodology is two-fold. On the one hand, we will consider results on graphs, find their (possibly many) formulations in terms of digraphs and see if and how they can be extended. Studying such extensions has been occasionally done, but the point here is to do it in a kind of systematic way. We will mainly focus on substructures in digraphs (complexity and conditions of existence) and extensions of graph colouring problems to digraphs. On the other hand, we will focus on the tools. We believe that many proof techniques have been too rarely used or adapted to digraphs and can be developed to obtain many more results. This in particular the case of median and cyclic order, the canonical decomposition of digraphs arising from matroids, the different notions of treewidth for digraphs, structural decomposition theorem, entropy compression and VC-deimension. Of course, those two approaches are not mutually exclusive but converge. Our goal is to develop the techniques to make advances in the above mentioned topics.

The development of algorithms for Autonomous Vehicles (AVs) faces important challenges throughout the design and implementation pipeline. The high cost and complex operation of real-world test-beds limits the experience an embedded Artificial Intelligence (AI) can gather, as it originates from a few vehicles that cannot be kept online extensively. For this reason, development often goes through a simulation stage or a testing step on a simplified system (e.g., smaller vehicles, standalone sensors or robotic models). In MultiTrans the focus is on the perception stage of AVs, which needs to provide a very accurate representation of the driving environment(s), that is used as an input for the following decision and control steps, while allowing a clear discrimination between similar but different contexts. The project takes the perspective of vision-based embedded systems (i.e., relying on cameras or similar sensors) that are among the most promising perception solutions. Their underlying sensing technologies however make them sensitive to an important research challenge: facing adverse conditions (such as bad weather or sun glare). In addition, knowledge transfer between different (real or virtual) environment suffers from two additional issues: reality gap, when a simulation/model fails to capture all the particularities of a real system, and the extended development time caused by the inherent repeated iterative process of adapting an algorithm from a system/domain to a different one. In MultiTrans, we propose to address these research issues by tackling autonomous driving algorithms development and deployment jointly. The idea is to enable data, experience and knowledge to be transferable across the different systems (simulation, robotic models, and real-word cars), thus potentially accelerating the rate an embedded intelligent system can gradually learn to operate at each deployment stage. The research hypothesis acting as a starting point of MultiTrans corresponds to the current state of deployment of autonomous driving technologies: AVs can be programmed (or are able to learn) to react and operate in controlled (or restricted) environments autonomously. The focus of our proposal is on the AI-side : research is needed to help these systems during the perception stage, enabling AVs to be operational and safer in a wider range of situations. The project is expected to contribute to substantial advances with respect to state of the art, by resulting in (i) A novel theoretical framework and new algorithms on transfer and frugal learning in virtual and real environments; (ii) Advances in multi-domain and multi-source computer vision for semantic segmentation and scene recognition applied to safe autonomous driving and (iii) The development of a robotic autonomous vehicle model demonstrator combined with a virtual world model. The novelty in this project is to develop an intermediate environment that allows to deploy algorithms in a physical world model. This additional step will allow to re-create more realistic use cases that would contribute to a better, faster and more frugal transfer of perception algorithms to and from real autonomous vehicle test-beds. This robotic platform will also enable to lead research focusing on multi-domain and multi-actor transfer by reducing the time and efforts required to build relevant use cases and multiple variants of these scenarios, thus allowing to achieve domain generalization. We will also explore frugal learning techniques such as few-shot learning would reduce the amount of samples require for the recognition/segmentation tasks to converge before transferring them. Thanks to the platform, we will be able to evaluate solutions for complex configurations in the virtual environment and then transfer them on the platform, bridging the gap between behaviour cloning (through imitation learning) and simulation.

Agronomy and biodiversity shall address several major societal, economical, and environmental challenges. However, data are being produced in such big volume and at such high pace, it questions our ability to transform them into knowledge and enable, for instance, translational agriculture i.e., rapidly and efficiently transferring results from agronomy research into the farms (“bench to farmside”). Semantic interoperability enables data integration and fosters new scientific discoveries by exploiting various data acquired from different perspectives and domains. D2KAB’s primary objective is to create a framework to turn agronomy and biodiversity data into –semantically described, interoperable, actionable, open– knowledge, along with investigating scientific methods and tools to exploit this knowledge for applications in science and agriculture. We will adopt an interdisciplinary semantic data science approach that will provide the means –ontologies and linked open data– to produce and exploit FAIR (Findable, Accessible, Interoperable, and Re-usable) data. To do so, we will develop original approaches and algorithms to address the specificities of our domain of interests, but also rely on existing tools and methods. D2KAB involves a multidisciplinary (and international) research consortium of three computer science labs (UM-LIRMM, CNRS-I3S, STANFORD-BMIR), four bioinformatics, biology, agronomy and agriculture labs (INRA-URGI, INRA-MaIAGE, INRA-IATE, IRSTEA-TSCF), two ecology and ecosystems labs (CNRS-CEFE, INRA-URFM), one scientific & technical information unit (INRA-DIST), and one association of agriculture stakeholders (ACTA). The consortium’s expertise ranges from ontologies and metadata, semantic Web, linked data, ontology alignment, knowledge reasoning and extraction, natural language processing to bioinformatics, agronomy, food science, ecosystems, biodiversity and agriculture. The project is structured with three work-packages of research and development in informatics and two work-packages of driving scenarios. WP1 will focus on ontologies/ vocabularies and turn the AgroPortal prototype into a reference platform that addresses the community needs and reaches a high level of quality regarding both content and services offered e.g., SKOS compliance, semantic search over linked data, text annotation, interoperability with other repositories. WP2 will focus on the critical issue of ontology alignment and develop new functionalities and state-of-the-art algorithms in AgroPortal using background knowledge methods validated in ag & biodiv. WP3 will design the methods and tools to reconcile the scenarios' heterogeneous ag & biodiv data sources and turn them into linked data within D2KAB distributed knowledge graph. It will also investigate exploitation of this graph through novel visualization, navigation and search methods. WP4 includes four interdisciplinary research driving scenarios implementing translational agriculture. For instances, an ontology-driven decision support system to select the most appropriate food packaging or an augmented semantic reader for Plant Health Bulletins. We will provide a unique scientific knowledge base for wheat phenotypes and offer the first agricultural data resource empowered by linked open data. WP5 will develop semantic resources for the annotation of ecosystem experiments data and functional biogeography observations. A plant trait-environment-relationships study will be conducted to understand the impacts of climatic changes on vegetation of the Mediterranean Basin. Within a dedicated work-package, we will focus on maximizing the impact of our research. Each of the project driving scenarios will produce concrete outcomes for ag & biodiv scientific communities and stakeholders in agriculture. We have planned multiple dissemination actions and events where we will use our driving scenarios as demonstrators of the potential of semantic technologies in agronomy and biodiversity.

Fluids, lithology and structure are known to play a major role in shallow fault slip modes but there is a major issue: Does one of them take precedence over the other two or are they linked? Are really fluids a key parameter? In recent years, studies in Japan, Costa Rica or New Zealand try to find the processes that control slip behaviours through marine campaigns and IODP drillings. The Ecuadorian margin is an exceptional laboratory to continue this international effort by adding a subduction zone to the list of those with contrasted slips at shallow depth. To go further in understanding a zone of the fault that could potentially participate to coseismic slip and create tsunami earthquakes, we need to estimate the in-situ scale of structure, fluids and lithology heterogeneities and identify their respective role. Thanks to three campaigns at sea and years of onshore/offshore data acquisition, this FLUID2SLIP project will participate to determine the exact role of fluids on slip behaviour around the updip part (0-15 km depth) of the seismogenic megathrust fault by localizing fluids and seismicity, imaging fault properties and deformation. To reach this ambitious objective, we designed 5 Tasks to : 1) quantify the fluid content and lithospheric structure by leading edge 2D and 3D seismic methods and heat flow mapping of physical properties of rocks, 2) characterize fluids flowing through the margin by analyzing their geochemical signature and relating fluid seepages to crustal deformation and interplate slip behaviour, 3) detect and locate seismic and transient signals activity with regard to fluids and slip behaviour using deep-learning techniques, 4) monitor seismic velocity time changes by ambient noise interferometry 5) model shear strength and slip processes in 4D onto two subduction zones in Ecuador and New Zealand and see if generic parameters/characteristics can be found. Our consortium gathers geoscientists and data sciences scientists from five French laboratories (Géoazur, I3S, Ifremer, ISTeP and Géosciences Rennes) and four international laboratories, covering thus a very broad range of expertise in seismic imaging, thermal structure, machine learning, seismicity, transient signals, marine tectonics and fluid seepage, and mechanics modelling. We can achieve our objectives in particular through a rare opportunity to have three planned cruises in 2022, each designed to reach specific goals. In addition to the deliverables to the scientific community (datasets, 3D geo-model, fluids database, 4D mechanical models), a science outreach sixth task, specifically addressed to the Youth, will be our contribution to the disaster risk reduction through the commitment of our Ecuadorian partner.

Companies are increasingly outsourcing their IT to cloud as a result of the attractive prices generated by resource pooling. Virtualization is used in the cloud to isolate applications, each being executed in a virtual machine (VM), which may be seen as an operating system (OS) with a limited amount of allocated resources. Power consumption is a major concern in the cloud (50-70% of the investment). VM consolidation, the most important mean for reducing the power consumption, is limited by virtualization which can run a VM exclusively on a single physical machine (PM). The latter must have enough available resources. This constraint leads to datacenter fragmentation, hardly addressable by consolidation considering the incongruous sizes of VMs. Because of these limitations, we have observed a significant level of resource waste in a cluster of our cloud partner (Eolas - Grenoble, France). The Scalevisor project proposes a virtualization that meets the nowadays cloud requirements: that is the unification of the ressourcesof several PMs into a single powerful PM. We propose to do this at the granularity of a rack where fast network links (Infiniband) are used.