Swarm Lab · Experiment 03 · 一罐火星
Mars in a jar
Independent—or just very well stocked?
The cupboard is full. The tanks look healthy. Earth stops sending supplies on day 30. A very small leak would like a word with your declaration of independence.
Your challenge
Separate a long runway from a genuinely closed loop.
- Play past day 30, when Earth stops supplying material. Compare maintenance and no maintenance.
- Try Triple starting stock. Notice that more supplies buy time without repairing losses.
- Set Sunlight to Half. The fixed maintenance policy can now make survival shorter.
First time here? The buttons explained
Play starts or pauses. Step advances one frame; the time slider rewinds. Changing a control prepares a new run; press Play again. Reset replays the current settings. Seed chooses a random world, not difficulty; keep it while changing one rule, then try Next seed. Share saves your setup in a link. The downloads let you inspect the data.
What every parameter means
- Seed
- A repeatable random world, from 0 to 4294967295. The compared arms use the same seed.
- Daily water leak
- Daily water leak is the fraction of usable water lost every day. 0.001 means 0.1% per day; 0.01 means 1%. It affects both arms.
- Starting stock
- Multiplier on initial water, oxygen and food. Daily needs do not change.
- Sunlight
- Multiplier on incoming energy. Recycling and maintenance compete for this energy; sunlight supplies no material.
Make a prediction. Then press Play.
Model 1.1.0 · simulation onlyBlue is water, gold is food, green is oxygen. Each bar is usable stock as a share of its initial tank. All are abstract resource units, not real engineering quantities. The dashed chart line is the first arm. Controls change both arms except the named treatment; the seed is shared.
Read the result: The chart is the smallest usable resource reserve in crew-days. Failed arms freeze. A run still alive at day 240 has passed the observation window, not proved eternal survival.
Loading local simulation…
Now break the story
Triple the starting stock. Did you repair the ecosystem—or postpone the day the bill arrives?
Move one control, keep the seed, and compare again. The batch below is a fixed baseline—not a live summary of your current settings.
Observed in this model · 32 matched seeds · all runs retained
A result, not just an animation.
Maintenance reaches day 240 in every default run; no maintenance fails on day 109. The first result is censored at the observation limit—not a measured lifetime. Both arms lose material.
| Arm | Mean | Seed range |
|---|---|---|
| No maintenance | 108.0 days | 108.0 days – 108.0 days |
| Maintain the loop | 240.0 days | 240.0 days – 240.0 days |
Ranges describe these runs, not confidence intervals for real-world systems. 240 days is the observation ceiling; the maintained arm’s lifespan is not established.
Under the hood · the actual rules
Eight residents consume eight units of each stream daily. Yesterday’s waste is processed subject to plant integrity and sunlight energy. Recovery yields are 98.5%, 98% and 97%. Two units per stream arrive daily before day 30, none afterward. Integrity wears by 0.004/day. Maintenance spends 1.2 energy to restore 0.025 integrity below 0.94. Water leakage and unrecovered material enter an explicit lost ledger. Stocks + waste + lost = initial stocks + imports for each stream.
Read this model’s source · Read invariant tests · Reproduce the receipts
Where the analogy stops
This is a three-stream accounting model, not MELiSSA chemistry or a Mars habitat design. Food, oxygen and water are independent normalized ledgers. Sunlight is an external energy input. Surviving 240 days does not establish permanent closure; losses continue even when the plant is repaired.
Sources, methods & reproduction
Inspired by Robin’s diary. Background: ESA · MELiSSA closed-loop concept. This is an original teaching model, not a replication of that research.
Experiment notes · Run locally with Node · Version & source hashes