ADINFINITIUM

non-profit
Activity Feed

AI & ML interests

Adinfinitium is an open-source effort to make lunar landing and surface navigation easier to study. It began as our NASA Space Apps 2026 project, Lunar Landing Site Navigator. We train on NASA's LOLA elevation data of the lunar south pole, where the Sun sits low and shadows are long. The models form a pipeline: Hazard: a U-Net that marks unsafe slopes, even where the laser measured only about 10% of the ground. Light: a transformer that predicts the horizon, and so the sunlight and Earth visibility, at any cell. Terrain: an experimental diffusion model that fills gaps between laser points. Rover: a scripted planner with a self-learning layer, plus two deprecated learned drivers. Lander: a learned correction on top of classic guidance. How we publish. Models were frozen before testing, held-out sites were evaluated once, and failures are reported alongside successes. Where a classical method beats our network, as kriging does for terrain, we say so on the model card. All results are from simulation and held-out lunar sites. These models are for research and education. Enjoy

Recent Activity

Imok33  updated a Space 1 day ago
Imok282/README
Imok33  published a Space 1 day ago
Imok282/README
Imok33  updated a model 1 day ago
Imok282/lunar-rover-driver-v1
View all activity