
Adinfinitium Light / Skyline Transformer
Status: Final · License: Apache-2.0 · Domain: lunar south pole (NASA LRO LOLA 5 m)
Predicts the 360° horizon elevation (skyline) at a lunar surface cell. Physics then converts the skyline into Sun fraction and Earth visibility.
Model details
| Task | Skyline regression: horizon height in 360 compass directions per cell |
| Parameters | 38,279,809 |
| Architecture | SkylineFormer: circular conv stem along each compass ray → transformer across 360 directions → zero-initialised head. Prediction = physics baseline + learned correction (an untrained model equals the physics backup). width 64, n_down 3, d 512, layers 12, heads 8, k_steps 64, n_az 360 |
| Training | 3,258 of 6,000 planned steps (54%); best checkpoint by validation skyline error on Site11 |
| Checkpoint | light_skyline_big.pt · 153,175,919 bytes · sha256 673b6715…e11d3 |
| Wiring | Not wired to the rover (the rover uses exact physics ray-casting) |
Benchmarks
Skyline mean absolute error in degrees (lower is better), against the coarse ray-cast baseline.
| Test set | Ours | Coarse ray-cast baseline |
|---|---|---|
| Skyline MAE, Site23 + Site42 | 0.450° | 0.570° |
| Skyline MAE, fresh sites | 0.243° | 0.321° |
| Sunlit-hours error | 7.6 h | 10.5 h |
Lower error on every one of the 19 test windows. Per site: Site23 0.534° vs 0.675°, Site42 0.270° vs 0.342°.
Limitations
- About 227× slower than the baseline (about 7.4 s vs 0.033 s per window); half precision changed the output by up to 0.327° and was rejected.
- Truth and model both ignore terrain beyond 1,280 m.
- Finishing the planned 6,000 steps changed nothing real (−0.0014°), so the frozen 3,258-step checkpoint is the one released.
Usage
# project code: https://github.com/oasisk/adinfinitium-handoff
from worldmodel.light.model import SkylineStudent, load_checkpoint, pick_device
Dream-RSI: gated offline self-improvement
Dream-RSI lets a model keep improving after the main training ends, offline and safely:
- Find failures: fly or replay many missions and record where the current model fails.
- Dream harder: generate and replay those hard cases (failure-boosted scenarios) and retrain a candidate, warm-started from the current model.
- Gate: the candidate is scored on a frozen validation exam (fixed seeds, paired with the current model and the physics backup).
- Lander: accept only if the candidate loses no mission that the current model or the physics backup lands (zero regressions) AND improves total success.
- Planner and rover: accept only if the frozen validation metric improves by a margin.
- Repeat from whichever model won. Rejected candidates are discarded, and the shipped model never gets worse.
Test maps and the final-exam site are never used in Dream-RSI: training uses training maps, and gating uses a separate validation map.
Measured results (all logged rounds)
| Loop | Rounds accepted | Effect on its gate metric |
|---|---|---|
| Planner (PlannerSetNet) | 1 of 5 | Validation Brier 0.01543 → 0.01353 (−12.3%) |
| Lander pilot (Site11 gate, 1,024 paired missions) | 0 of 7 | Pilot unchanged; every candidate lost 12–25 missions, so the gate held |
| Rover brain v3.1 (self-learning on stressed worlds) | 9 of 36 | Blocked route improved at 4/4 sites, late light 3/4, dusty panels 1/4, early light 0/4 |
| Rover brain v3 | 9 of 72 | Blocked route improved at 6/6 sites, dusty panels 2/6 |
| Rover learned driver v2 | 5 of 38 | Returned rate stayed 0.0% |
Across all loops, 18 of 42 first rounds on a new problem were accepted, against 6 of 120 later rounds. What this shows: Dream-RSI reliably repairs specific failures (blocked routes, late light, model faults) without ever shipping a regression. In unchanged conditions it rarely finds a safe gain (0 of 32 rover rounds accepted).
Projections after more Dream-RSI (Projected, not measured)
Median after 20 more generations, from the planner's own Dream-RSI log:
- Planner Brier: 0.0145 → 0.0094 [0.0043, 0.0145]
Scenario only, for models where Dream-RSI has not been run yet (it borrows the improvement rate seen in the planner and rover loops):
- Hazard fresh F1: 0.757 → 0.779
- Light error: 0.450° → 0.410°
- Terrain error: 2.01 → 1.84 m
The lander and rover v3.1 are projected to stay unchanged, because their gates have not accepted anything in unchanged conditions.
Limitations and next steps
- The zero-regression gate is strict: it also rejected lander candidates that were slightly better overall (best: 991 vs 990 landings).
- Training rounds were short (4–5 updates) because of the compute budget.
- Next: harder self-generated scenarios, longer rounds, and a fresh unseen-map exam after each accepted generation.
Intended use and scope
Research, education and benchmarking of lunar landing-site navigation methods. Not for flight, safety-critical or operational landing decisions. All results are simulation or held-out-lunar-site results, not real-mission performance. Models were frozen (SHA-256) before held-out evaluation; each test site was evaluated once.
About Adinfinitium
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, and trains on NASA LRO LOLA 5 m terrain of the lunar south pole. The model chain is: hazard network → light network → terrain model → rover brain → lander pilot. See the whole family at https://huggingface.co/Imok282.
Source code, full records and the hand-off package: https://github.com/oasisk/adinfinitium-handoff (checkpoint SHA-256 values: CHECKPOINTS.sha256).
Citation
@misc{adinfinitium2026lunar,
title = {Adinfinitium open models for lunar landing-site navigation},
author = {Adinfinitium},
year = {2026},
howpublished = {Hugging Face, https://huggingface.co/Imok282},
note = {NASA Space Apps 2026, Lunar Landing Site Navigator}
}

