Europe's future in space is not only built on large missions and spectacular launches, but also on a series of bold, unconventional ideas that need to be quickly tested, validated, and turned into useful technologies. This is precisely what ESA's Open Space Innovation Platform (OSIP), as part of the Discovery & Preparation programme, does: it finds exceptionally good proposals, connects science and industry, and secures the first round of funding to turn ideas into tangible results in a short cycle. In the first half of 2024, OSIP supported a mosaic of 51 research and development activities, among which five projects particularly stand out, vividly demonstrating how Europe plans to fund the future of space – from foundational AI models for Earth observation and autonomous navigation for close encounters of spacecraft, to a huge "Moon-and-Mars" laboratory chamber, an electronic "pencil" for dust removal, and next-generation diamond quantum magnetometers.

Between January and June 2024, OSIP served as an entry point into ESA's innovation ecosystem for teams solving very specific challenges: how to better understand the planet given the flood of satellite data, how to safely manage spacecraft encounters in orbit, how to test rover platforms and fragile systems in conditions similar to lunar and Martian ones, how to deal with the abrasive electrostatics of lunar regolith dust, and how to fit highly precise magnetic field measurement into small, low-power space platforms. Below is an expanded overview of these five OSIP-driven activities, why they matter, where their commercial potential lies, and how they fit into the European space programmes that will define this decade.

OSIP as a fast track for ideas: how Europe is shortening the path from proposal to prototype

The Open Space Innovation Platform is designed as an "entry point" for unconventional concepts. Through open channels and thematic campaigns, researchers from academia, start-ups, and industry propose solutions that undergo a rapid technical evaluation. The most promising ideas move into one of three tracks: feasibility studies (to quickly check if they "hold water"), co-funded research (PhD and postdoc topics with a clear technological outcome), or early technology development (TRL leaps towards validation in a relevant environment). The model is deliberately "portfolio-based": risk is distributed, and decisions on scaling are made based on data and demonstrations – not just on good intentions.

In 2025, this approach becomes even more important. The global race for autonomous in-orbit servicing, the increasing density of satellites in low Earth orbit (LEO), plans for longer stays on the Moon, and sample return from Mars demand technologies that turn fragility into robustness: systems that are smaller, lighter, and more power-efficient, algorithms that work under the constraints of "space-grade" processors, and test facilities where dusty, cold, and ultra-vacuum environments can be replicated on Earth. OSIP catches such attempts early and directs them towards specific missions.

A foundational AI model for Earth observation: "any-sensor" learning and language-based control

A team from the École Polytechnique Fédérale de Lausanne (EPFL) is working on creating a foundational AI architecture that is sensor-agnostic and learns common representations from radar (e.g., Sentinel-1) and optical (e.g., Sentinel-2) images over the same areas. Instead of the classic "one sensor, one model" approach, the goal is for a single network to recognize patterns regardless of whether they come from SAR, multispectral cameras, or some future instrument. This bridges differences in spatial resolution, spectrum, and temporal frequency, and the results show that the model can effectively perform tasks even outside the dataset it was trained on.

The next step is to merge vision and language: the model should "understand" textual descriptions and answer questions about the content of a scene. This means a user could type a query like "find flooded areas along the Sava river in April" or "show changes in urban sprawl between 2018 and 2025" and get a meaningful, verifiable output. Such language interaction dramatically lowers the barrier to using Earth Observation (EO) data for non-specialist teams – from civil protection and crisis management to agriculture, energy, and investigative journalism.

Why is "any-sensor" important right now? Because constellations are growing, and missions complement each other: radar sees through clouds and at night; optics provide rich spectral content; hyperspectral instruments detect chemical signatures; altimeters and lidar provide geometric parameters. A unified latent representation of all these sensor "languages" shortens the time to insight and reduces operational costs. Additionally, standardized EO benchmarks increasingly favor models that generalize across geographies and modalities, confirming that this direction is the right one.

On the operational side, use cases with high social impact stand out: rapid damage assessment after floods and fires, landslide detection, monitoring of forest diseases and agricultural stress, mapping of urban heat islands, tracking coastal erosion, and combating illegal activities (illegal construction, logging, sand extraction). In such scenarios, language-based search shortens the path from question to answer, and "any-sensor" learning mitigates data gaps.

SpaceSite Lab: "Moon and Mars" in a large-diameter dusty, windy vacuum chamber

How hard is it to reconstruct the Moon and Mars on Earth? Hard enough to require a chamber about 30 meters in diameter and 7 meters high that combines a vacuum, extreme temperatures, wind, and mobilized dust. This is precisely what the Dusty-Windy Thermal Vacuum Chamber (DWTVC) concept envisions as part of the SpaceSite Lab, a joint venture between the Danish Technological Institute (DTI) and Aarhus University (AU). The goal is to create a full-scale test and research facility where rover platforms, robotic arms, mobile bases, and sensitive mechanisms can be tested under conditions as close to reality as possible – from the lunar polar regions with electrostatically charged regolith to the hot, thin Martian atmospheres.

Why is such a chamber strategic? Because it scales testing from laboratory samples to complete systems and scenarios. It can simulate "dust storms," measure the degradation of optical and thermal surfaces, verify the reliability of connectors and seals, and validate dust mitigation systems before they are sent on a spacecraft. A commercial component outside of space has also been conceived: measuring aerosol emissions from vehicles and furnaces, testing large ventilation systems, and even agricultural experiments under specific atmospheres. The conceptual designs are linked with leading architectural firms, and the technical configurations follow the needs of future European missions in the Terrae Novae programme and the Argonaut lander programme.

A full feasibility study is underway, covering technical, architectural, and infrastructural aspects, as well as the financial and shareholder map. This creates the prerequisites for decisions on phased construction and public-private partnerships. If Europe wants to seriously test large lunar and Martian systems, this kind of "terrestrial space" environment will accelerate development and reduce risk.

AI in service of close encounters: relative navigation for in-orbit servicing

In-Orbit Servicing (IOS) is slowly transitioning from a "vision" to demonstrations. However, for a "Chaser" to safely approach a "Target," a system is needed that can handle uncertainties: the shape of the target object is not fully known, its optical properties change, lighting conditions vary, and everything happens with limited processor resources and strictly controlled fuel consumption. A research team from the Polytechnic University of Milan (Politecnico di Milano), in partnership with ESA and industry, is developing a relative navigation algorithm with AI "in the loop" that is robustly designed for precisely these conditions.

Validation is being carried out on two levels. First, the algorithm is run on a space-qualified processor to verify its embeddability without dramatic changes to the spacecraft's architecture. Second, "camera-in-the-loop" and "processor-in-the-loop" tests are being conducted at the Thales Alenia Space facilities in Cannes with varying target appearances, testing its resilience to surprises in real operational conditions. The market potential is strong: from space debris removal to refueling and installing new modules on existing satellites, IOS is a segment that will require reliable, certified GNC systems with learning elements in the coming years.

In the long run, such algorithms could also be applied to robotics on the surfaces of the Moon and Mars, where there is no GPS and conditions change rapidly. A successful demonstration on "space-grade" hardware is thus a prerequisite for European autonomy in IOS – and an important safety mechanism for crowded orbits.

A "pencil" against dust: how a collimated electron beam removes regolith

Dust is the silent enemy of crewed missions and robotic systems. Fine, abrasive, and electrostatically charged particles get into seals, dirty optics, damage thermal surfaces, and reduce energy production on solar panels. The Romanian National Institute for Laser, Plasma and Radiation Physics (INFLPR) is investigating a dust removal technique using a collimated, pulsed electron beam with an energy of about 13 keV. The idea is elegant: the electrons transfer momentum directly to the dust particles, ejecting them from the surface without any prior preparation – no special coatings, embedded wires, or substrate adjustments – making the method more universal than many existing solutions.

The limits of safe and effective operation are currently being tested in the laboratory. The spot diameter (on the order of 10 mm), pulse duration (tens of microseconds), repetition frequency, and total current per pulse are carefully adjusted to achieve high cleaning efficiency with minimal risk of damaging the substrate. Since the technique is inherently suitable for low-pressure and vacuum conditions, applications on the Moon are obvious. But the potential extends to industries on Earth as well: rapid cleaning of sensitive optics, maintenance of solar farms in deserts, servicing of sensors in aggressive atmospheres – these are all markets that could benefit once the concept is turned into production prototypes. Plans include an international patent and commercialization through a start-up channel.

Diamond quantum magnetometers: NV-centers for vector measurement in a small package

Measuring magnetic fields in space is fundamental to studying the internal structure of planets, monitoring space weather, protecting electronics, and navigation. Classic magnetometers, however reliable, often suffer from limitations in mass, power consumption, and long-term stability, and vector measurement often requires multiple sensors and careful calibration. A team from Hasselt University is developing a miniaturized magnetometer that relies on the quantum physics of diamond – on so-called NV (nitrogen-vacancy) centers whose quantum state is excited and read out optically, with microwave control. This approach offers a wide dynamic range, robustness, and the ability for vector measurement in a compact form.

The concept has already experienced space through the OSCAR-QUBE student mission on the International Space Station, which has paved the way for a "second generation" addressing limitations and taking steps towards "on-chip" integration. The goal is an instrument small and power-efficient enough to fit on nanosatellites and in constellation architectures, with the long-term stability and accuracy needed for scientific and operational missions. Prospects beyond space include geological surveys, medical devices, and industrial solutions that require precise, stable, and rapid magnetic field measurement.

From idea to impact: where the projects fit into Terrae Novae and Argonaut

The OSIP-driven activities are very concretely linked to future European plans. The Terrae Novae programme requires resilient technology for the dusty polar regions of the Moon – without a solution for dust mitigation, everything is at risk: from mechanical assemblies to power and communication. At the same time, the Argonaut lander programme requires components tested in a relevant environment before they are sent on missions that carry the weight of an entire continent. Full-scale test chambers and dust "cleaners" are therefore not a luxury, but a prerequisite for reliability.

In orbit, however, another pressure is growing: congested LEO and MEO pathways require servicing, inspections, and safe debris removal. AI-powered navigation that is resilient to uncertainties and provable on "space-grade" hardware is becoming a kind of standard that every serious IOS mission will require. On Earth, foundational "any-sensor" AI models are becoming a common work platform for public services, researchers, and companies that want to get from EO data to fast, verifiable decisions.

What happens after the first half of 2024: the rhythm of demonstrations in 2025

Although the focus here is on the January–June 2024 period, development intensified during 2025. The research community is publishing new architectures for multi-sensor foundational models, discussing the standardization of comparisons and the balancing of datasets across geographies and modalities, and accelerating the transfer from the lab to operational workflows. Industry, spurred by European and national programmes, is preparing demonstration flights for IOS, and academic and industrial teams continue their search for more durable dust mitigation solutions. The common denominator is the same: faster and more responsibly from idea to impact.

Quick guide: official sites and useful links

• OSIP & Discovery – how to apply, what channels exist, and what campaigns are being run: OSIP – ESA.


• SpaceSite Lab (DWTVC) – summary of objectives, scope, and study status: SpaceSite Lab.


• AI for relative navigation – research description and test plan on a "space-grade" processor and optical-robotic loops: Autonomous AI-aided Relative Navigation.


• Electron beam for dust removal – the "pencil" concept and key pulse parameters: Electron-beam Dust Mitigation.


• OSCAR-QUBE & NV magnetometer – the diamond quantum sensor in space and the path towards miniaturization: OSCAR-QUBE.