ERC Grant to bring self-healing AI hardware into space

ERC Grant to bring self-healing AI hardware into space

The GROW-AI team, led by Professor Sebastian Risi at the IT University of Copenhagen, has been awarded a 2025 ERC Proof of Concept Grant for a project aimed at making future space technologies more resilient to radiation and hardware failures.

Researchartificial intelligence

Written 6 July, 2026 07:27 by

How do you build computer systems that can repair themselves when faults occur? That question lies at the heart of a new research project at the IT University of Copenhagen, which has received a 2025 ERC Proof of Concept Grant. The grant will help Professor Risi and his team turn findings from the earlier ERC-funded GROW-AI project into technology with the potential to transform the way hardware is designed for space applications and other critical systems.

For us, the ERC Proof of Concept Grant means that we can now take something that has so far existed primarily in simulations and begin to turn it into real technology. It is an opportunity to develop the research into something that could be applied in practice," says Professor Sebastian Risi.

The aim is to develop a new type of AI-powered hardware that can automatically adapt and restore its functionality when faults occur. Such technology is particularly relevant for space applications, where satellites, rovers, and other systems are constantly exposed to radiation that can damage electronic components.

When a Single Fault Can Have Major Consequences

Modern spacecraft rely on advanced computer systems to collect data, navigate, and perform complex calculations. However, in space, cosmic radiation can alter individual bits in electronic circuits. Even minor faults can, in the worst case, cause systems to malfunction or fail completely.

Today, this challenge is typically addressed through additional hardware, protective materials, or the manual reprogramming of systems. However, these solutions increase weight, energy consumption, and resource requirements**, and** they cannot necessarily address every type of fault.

"Space is a harsh environment. Computer systems in spacecraft are constantly bombarded by radiation. Today, this is typically addressed by building additional protection and redundancy into the systems. Our approach is different because we are trying not only to prevent faults but also to create hardware that can respond to them and reorganise itself," says Professor Sebastian Risi.

The research group, which includes Professor Sebastian Risi, Postdoctoral Researcher Eleni Nisioti, and PhD Student Marcello Barylli, is therefore exploring new ways of creating more resilient systems. "This grant simply would not have happened without the GROW-AI team, especially Eleni and Marcello," says Professor Sebastian Risi. "Marcello's proof-of-concept results convinced us that the idea could work in practice, and Eleni's scientific leadership has been essential to bringing the whole project together."

Inspired by Nature's Ability to Adapt

The project builds on findings from GROW-AI, through which the research group developed so-called Neural Developmental Programs. The method is inspired by biological processes, in which complex organisms emerge from simple rules while adapting to changes in their environment.

Rather than viewing an electronic circuit as a fixed structure, the researchers regard its individual components as small units that can work together to find new solutions if parts of the system are damaged. When a fault occurs, the components are designed to exchange information locally and reorganise, allowing the system to continue to perform its intended function.

"We are developing computer chips that can, in a sense, heal themselves. When a conventional chip is damaged, it often simply stops functioning. Our idea is to create systems that can instead adapt to damage and continue to operate," explains Postdoctoral Researcher Eleni Nisioti, one of the project's lead researchers.

From Artificial Intelligence to Self-Healing Hardware

The researchers aim to apply artificial intelligence directly to programmable chips known as FPGAs (field-programmable gate arrays). FPGAs are already widely used in space technologies and high-performance computing systems because they are flexible and energy-efficient.

This new approach differs from traditional methods because the system does not need to rely on external intervention or extensive reprogramming. Instead, the chip itself is designed to detect faults and respond by reconfiguring its functions.

If the technology performs as expected, faults could be addressed more quickly and using significantly less energy than with existing solutions. The researchers describe the project as a step towards hardware that more closely resembles biological systems: resilient, adaptive, and capable of responding to unforeseen events.

Next Step: From the Laboratory to Real-World Applications

With the ERC grant, the research group can now move from simulations to testing physical hardware. The project includes software development, the training of AI models, and experiments using real FPGA devices.

At the same time, the researchers are collaborating with Neurospace, a company developing AI-based space technologies. In the longer term, the ambition is to test the technology in partnership with the European space sector.

The path from basic research to practical application is a long one, but facilitating that transition is precisely the purpose of the ERC Proof of Concept programme. It gives researchers the opportunity to explore how they can translate promising research findings into solutions with societal and commercial value.

The broader perspective extends beyond making space technologies more efficient. It is about exploring a new way of thinking about hardware - one in which machines not only execute instructions but can also adapt as the world around them changes.

"Space is an obvious place to begin because hardware in that environment is exposed to radiation and is extremely difficult to repair. In the longer term, however, the technology could also prove relevant for large-scale data centres, where even minor faults can have significant consequences when they occur at scale. Another exciting prospect is neuromorphic hardware - computer systems inspired by the structure and function of the brain. If this type of hardware becomes more widespread, I believe self-healing and adaptive capabilities could become increasingly important," says Postdoctoral Researcher Eleni Nisioti.

Potential Far Beyond Space Applications

Although space technology is the initial application area, the ambitions extend far beyond it. Many emerging technologies depend on hardware that must operate reliably under challenging conditions, including autonomous systems, neuromorphic computing, quantum technologies, and critical infrastructure.

What these fields have in common is the need for systems that can continue to function even when parts of the hardware fail. If the researchers can demonstrate that their approach works in practice, the technology could have implications far beyond the space sector.

"If successful, the hardware of the future could become less centralised, more resilient, and far more organic in the way it operates. Instead of rigid systems, we could have computers that reconfigure themselves and maintain their functionality even when something goes wrong," says Professor Sebastian Risi.

The ERC grant is worth approximately DKK 1.2 million, and the project will begin on 1 July 2026 and run until the end of 2027.

Find out more about GROW-AI here

Further information

Jari Kickbusch, phone 7218 5304, email [email protected]