Adinfinitium

Adinfinitium Rover Driver v2 (imitation / DAgger) β€” DEPRECATED

Status: Deprecated (failed; research record) Β· License: Apache-2.0 Β· Domain: lunar south pole (NASA LRO LOLA 5 m)

⚠️ DEPRECATED MODEL. This checkpoint did not meet its pre-registered success criteria and is superseded by the scripted lunar-rover-brain (v3 / v3.1). It is kept online only as a transparent research record of a negative result. Do not use it for planning or navigation.

Same DriverFormer network as v1, trained by imitation with DAgger from six planner variants. Pre-registered verdict: failure. Published as a research record.

Model details

Task Rover driving policy (imitation of scripted planners)
Parameters 11,377,354
Architecture DriverFormer (as v1): d 384, layers 6, heads 8, crop 64, stem 64
Training DAgger imitation; 105 episodes, 14,605 steps, 110 minutes over 32 training worlds; selected on validation Site11 (candidate 6 of 9)
Checkpoint rover_driver_v2.pt Β· 45,543,277 bytes Β· id 60985f9de2 Β· sha256 cecdf244…79375

Benchmarks

8 held-out worlds (Site23 and Site42 Γ— 4 dates), simulation only. Pre-registered verdict: FAILURE.

Imitation driver v2 vs naive planner

Reached ice Returned Died
This policy 7.4% 1.2% 95.3%
Naive planner 19.6% 9.3% 86.8%

On Site11 (validation) it returned 26.8% against the planner's 40.7%.

Limitations

  • It memorised routes: it agrees with the planner on 86–93% of return-leg steps on trained terrain but only 15–26% on unseen terrain.
  • Site11 was an easy validation site; training site Site04 is 99.5% over the slope limit and taught nothing.
  • Dream-RSI on top of it accepted 5 of 38 rounds, but its returned rate stayed 0.0%.

Usage

Not recommended. Loader for research: worldmodel.rover.policy.load_checkpoint (project code on GitHub).

Dream-RSI: gated offline self-improvement

Dream-RSI lets a model keep improving after the main training ends, offline and safely:

  1. Find failures: fly or replay many missions and record where the current model fails.
  2. Dream harder: generate and replay those hard cases (failure-boosted scenarios) and retrain a candidate, warm-started from the current model.
  3. 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.
  4. 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)

Dream-RSI rounds accepted per loop

Loop Rounds accepted Effect on its gate metric
Planner (PlannerSetNet) 1 of 5 Validation Brier 0.01543 β†’ 0.01353 (βˆ’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 β†’ 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 β†’ 0.779
  • Light error: 0.450Β° β†’ 0.410Β°
  • Terrain error: 2.01 β†’ 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 β†’ light network β†’ terrain model β†’ rover brain β†’ 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}
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Video Preview
loading