Introduction, objectives and background

 In the context of rapid changes of our planet Earth, explosive population growth, accelerating industrialization, especially in the emerging countries, increasing exploitation of resources, serious natural hazards, recent natural disasters, and, importantly, the growing number of complex systems interacting with their environments, the European Training Network (ETN) MENELAOS^NT addresses the ever more significant problem of scientifically and technologically exploring the world on micro- and macroscopic scale. The research and development goal is to provide scientists with deeper insights, better understanding and more information to understand and monitor the basic processes and environments in order to predict and possibly control critical evolutions. This would support decision makers with more sophisticated and reliable information for deciding on sustainable measures.

MENELAOS^NT addresses societal key challenges, e.g., sustainable agriculture and forestry, bioeconomy, environmental changes, resource efficiency, as well as protecting freedom and security of the European society.
MENELAOS^NT applies Novel Technologies to realize multimodal – multi sensor data fusion to optimally combine the information, delivered by different sensors (in-situ/remote, optical/non optical) on different scales, with different resolutions and with different reliability. Traditional remote observation, as well as local (in-situ based) approaches have demonstrated to be insufficient to cope with the ample but very tailored information need of the decision-making bodies.

The continuously increasing sensitivity, resolution and accuracy of remote sensing on the one hand, and the steadily growing sensor range and increasing field of view of short and medium range 2D/3D-sensorics, especially when combined with mobility, on the other hand, are about to close the classical gap between analysing unknown scenarios from outside (remote) and from within (in-situ), however, at the cost of exponentially increasing data volume. The “Big Data” approach of acquisition and collection of all possible kinds of data appears feasible and popular, but comes at the cost of exponentially growing data sets. In order to promote sustainable and resource conscious information generation, the technical state of the art of analogue to digital conversion, followed by denoising, interpreting and extracting information must give way to directly acquire information instead of data. The novel discipline of Computational Sensing describes the capacity of a sensor or exploration system to directly acquire focussed information rather than simply data, exploiting the dramatically exploding efficiency of computer-based signal processing approaches. The most prominent representative of this discipline is the novel mathematical theory of Compressed Sensing (CS) and Compressed Learning (CL): Rather than first measure everything and compress afterwards, one of the core ideas is to develop new technologies to directly acquire compressed information and fuse these very different sources of information to derive the specific information, needed to explore, understand and learn in and from a specific environment.

Research methodology and approach

MENALAOS^NT research is structured into four main work packages addressing the different hierarchical stages and abstraction levels.

1. Novel Sensors and Systems (WP 1) for image and information generation. Sensors constitute the very first and elementary stage in an information system. They exploit different modalities and different sensor effects, deliver measurements, which contain more or less information depending on the individual signal to noise ratio.
2. Sensor Information Flow and Integration (WP 2) with the novel paradigm of Analogue to Information Conversion (AIC). As opposed to classical sampling and converting from analogue to digital, AIC aims at integrating many of the CS principles of WP 3 already in hardware as part of the sensor interface thus reducing data rates or increasing information rates.
3. Novel Approaches to Sensor Data Processing (WP 3) exploiting the novel mathematical theory of Compressed Sensing/Compressed Learning (CS/CL) for the engineering and information retrieval applications of this ETN. In simple terms, the mathematical theory of CS states that even a small number of measurements obtained via sub-Nyquist sampling can guarantee stable and even exact information reconstruction under certain conditions. Investigating the limiting conditions, the feasibility, the applicability and perspectives of these principles at each stage of a sensor information system constitute a central issue throughout the ETN.
4. High Level Information Mining (WP 4). This work package addresses the empirical observation that sensor data is not identical with information, due to the problem that information is seldom directly observable and very often “disguised” in measurement noise. Information extraction implies learning from measurement data giving rise to all techniques of machine learning, deep learning, especially in combination with data compression, giving rise to compressive deep learning. Information furthermore depends on how specific questions are formulated and is finally subject to the perception of individuals. High level information mining closes this apparent (sometimes called semantic) gap.

Integration of innovative research into innovative training

1. AMO is a manufacturer of nanotechnological optical and nano-optical 2D raw sensors, directly interfacing with pmdtec and CiTIUS working in optical ToF imaging, but using different modulation schemes (continuous wave CW, or pulse modulation PM) and with FHR (non-optical sensing) and with ZESS (optical and non-optical (SAR) sensing), however, using different sensing setups and geometries. Also, they adhere to different signal processing “schools” leading to complementary (and also competing) approaches. CiTIUS, ZESS, CEOSpace Tech, AMO, and pmdtec all work on novel types of optical 3D imaging sensors, exploiting, however, different technological base principles.
2. ZESS, DLR, FHR, SPIS, CEOSpace Tech, SAMPL, and GAMMA all work in SAR-based (non-optical) remote sensing, focussing on different stages of the overall sensing and processing chain.
3. SPIS, ZESS, SAMPL, FHR, and CiTIUS do research and development on information-flow-optimized and highly integrated sensor pre-processing hardware.

• A central research interest of all network partners are novel approaches to sensor data processing. Here ZESS, SPIS, FHR, SAMPL and CEOSpace Tech jointly exploit the methodological framework of compressive sensing, putting their individual research foci on different methodological flavours and applications.
• Finally, CEOSpace Tech, DLR, SPIS, SAMPL, INSITU, and GAMMA very intensively research and apply high level infor-mation mining, constituting the methodological framework for all application-directed work of the partners and multiplying the impact by interfacing with a broad user spectrum.

Excellent Career Perspectives for doctoral students (ESRs)

Sensors and sensor systems for environmental observation, exploration and monitoring, including novel techniques of information acquisition, processing and digesting constitute a key challenge and form a key future market with a huge potential and with attractive and challenging career opportunities in science and industry. MENELAOS^NT will interlink international key players, scientific experts, their institutions and the ESRs to work together in innovative training environments giving the ESRs access o an existing, well-functioning international network of universities, excellent research centres and industry partners, providing them with an ecosystem of advanced research environments. The obvious exponential growth of sensors and sensing applications implies excellent career perspectives for well trained experts being able to disseminate and integrate such an explosion of technical and technological enablers.

MENELAOS^NT fellows will be trained by research in their individual projects, they will receive high level dedicated training by specialised courses leading them to the core scientific questions of the projects. MENELAOS^NT fosters new generations of (younger) scientists, following the EU Principles on Innovative Doctoral Training, namely research excellence, attractive institutional environment, the triple “i” dimension (international, intersectoral, and interdisciplinary), transferable skills training, and quality assurance at the highest possible level.

By integrating key players in remote sensing and earth observation, such as FHR, DLR, and GAMMA the ESRs get in-volved with very extremely advanced sensing infrastructure and its further development. FHR and DLR represent two of the largest research institutes with international standing covering the field of sensors, sensor data processing and information mining with a focus on remote sensing/remote exploration systems. DLR, FHR and GAMMA – with their extremely large user network of TerraSAR-X, TanDEM-X, as well as optical remote sensing data – will serve as interfaces and multipliers for ex-tremely broad user communities deepening the cross sectoral training. SAMPL, SPIS, CEOSpace Tech, FHR and ZESS are key players in all fields of sensor signal processing, information extraction and high level information fusion, providing the ESRs with an optimal training backbone on sophisticated data processing and information extraction technology.

By integrating companies like AMO, pmdtec, INSITU and GAMMA with their short time to market periods and high innovation potentials, ESRs directly experience the thrilling spirit of research, put to practice and tested under real-life conditions as well as novel products and services. In fact, INSITU has been awarded as the Best Small Company in the region of Galicia (Spain) in terms of Transfer of Technology, in recognition of the advances made in terms of geographic 3D big data.

Dual Supervisor Support/Supervision Strategy (D4S)

The network partners can build upon established international postgraduate education infrastructures. They explicitly adhere to the European Charter for Researchers. Combining the best practice experiences, for each ESR the Dual Supervisor Support/Supervision Strategy (D4S) includes:

• Joint supervision of each ESR by two scientific supervisors from different organizations, regularly (every two weeks) and on demand
• Individual career development plans (CDPs) set up and signed with the two supervisors ensuring optimized doctoral training and scientific support (CDPs will contain a binding schedule for regular meetings with the ESRs’ supervisors and advisors (D4S) for scientific exchange, progress monitoring and trouble-shooting)
• Embedding of each ESR into the working groups of their supervisors invoking the group leaders (senior researchers)
• Invoking scientific advisors at the secondment hosts already in the initial phase of each ESR
• Exposure to industry and other relevant employment sectors
• Scientific quality assessment by appropriate dissemination of individual research results (project reports, conferences, journal papers, annual fall presentations, midterm summer school)
• Social and personal support ensured by programme infrastructure
• Continuous improvement of support quality by incorporating feedback from the ESRs