IMPOSTER

The IMPOSTER (rendering IMages for oPtical navigatiOn toSupport arTificial intElligence tRaining) project develops a framework for generating highly realistic synthetic image data to train AI models for visual navigation in space missions. The sandbox software aims towards an AI-based, explainable, and uncertainty-aware pose estimation algorithm  in collaboration with The Exploration Company. This enables navigation toward non-cooperative objects like space debris, supporting satellite servicing and removal missions. It also features a modular architecture with interchangeable target objects and camera models, enabling parametric scene modifications. It optimizes rendering through CAD-to-3D conversion, balances detail with realism, replicates lighting and materials for high-fidelity output, and supports AI applications with reliable training data and automatic labeling.