The fields of artificial intelligence (AI) and extended reality (XR or immersive technologies) are generating a multitude of innovative space applications. Examples include game engines, used in the video game industry, and their novel use in creating realistic simulations for data capture or next-generation convolutional neural networks (CNN) for computer vision of Mars images. However, these technologies are severely limited in terms of availability of data for training or testing. This is the case, for instance, for the Mars Sample Return Mission, which plans to combine AI and XR to simulate datasets that facilitate the estimation of the pose of objects on the red planet.
The European Space Agency (ESA) is dedicating efforts and funding to solving this problem through AI and XR, because of the enormous potential of these tools to change the way the international community interacts with space.
Within this framework, the project ‘AI-Aided-XR: AI Aided eXtended Reality applications’ has just concluded, developed by an international consortium with the participation of the ARTEC group, from the Institute of Robotics and Information and Communication Technologies of the UV. AI-Aided-XR addresses the lack of data in AI space applications through a simulated environment generated by fractal algorithms. The role of the University of Valencia team in this initiative has been to recreate lunar surface terrains as a study model that can be extrapolated to celestial bodies in general.
“Simulating synthetic soils of celestial bodies allows us to train the navigation algorithms used in the rovers that explore the surfaces of other planets”, explains Marcos Fernández, researcher at IRTIC, head of the ARTEC group and of the UV's collaboration in the project. “The synthetic environments we develop must be based on real data. For this, and using artificial intelligence, we learn the fractal values and real characteristics of the lunar areas of interest, which will then allow us to generate variants with the same random patterns”, adds the researcher. “The first phase of exploration of this hybrid technology has been a success and we hope that ESA will extend the work for use in real missions and on other celestial bodies”, he concludes.
Funded by ESA, the project has been developed by an international consortium coordinated by the GMV group (GMV NSL Ltd, from the United Kingdom, GMV Soluciones Globales Internet SAU, from Spain and GMV Innovating Solutions Sp, from Poland), ESA’s European Space Operations Centre, in Germany, and the University of Valencia.