Physarum Slime Mold Simulator
Thousands of single-cell agents lay pheromone trails, follow each other's strongest scent, and self-organise into living networks. The same algorithm a real Physarum slime mould used to rediscover the Tokyo subway map.
Palette
What is happening on the canvas
Every white-ish trail is the cumulative path of thousands of tiny agents. Each agent follows three steps every frame:
- Sense the pheromone field at three sensor points — front, front-left, front-right.
- Turn toward whichever sensor reads strongest.
- Move forward one step and deposit a little pheromone where it lands.
In parallel, the pheromone field diffuses to neighbouring cells and decays a few percent per frame. There is no central planner — but the agents constantly reinforce the trails that other agents have already taken, and the system condenses into networks of veins, loops, and hubs.
The parameter zoo
| Parameter | Effect |
|---|---|
| Sensor angle | Wider angle finds trails further off-axis — fuzzier branches. |
| Rotate angle | How sharply agents turn each step. Equal to sensor angle gives stable lines. |
| Sensor distance | How far ahead they look. Short = thin, dense webs; long = sweeping curves. |
| Decay | High decay → fragile, fast-changing networks; low decay → persistent veins. |
| Agent count | Density of the resulting network — more agents form thicker bundles. |
The classic Jeff Jones (2010) settings live around sensor angle 22°, distance 9 px, rotate 45°, decay ~5%. Push values around and you can reproduce neuron-like nets, blood-vessel patterns, or featureless noise.
A real slime mould that solves maps
Physarum polycephalum is a real organism with no nervous system, just one giant cell with many nuclei. In 2010 researchers placed oat flakes at the positions of Tokyo's major cities on a wet plate. The slime mould grew between them and pruned its tubes — within hours, it reproduced a network strikingly similar to the actual Tokyo subway map, just from local rules.
The algorithm in this simulator is a digital model of that behaviour. The same patterns turn up in fungal mycelia, river deltas, and the cytoskeleton — anywhere transport networks have to be built without a blueprint.
Things to try
- Crank deposit + lower decay for thick, glowing veins that look biological.
- Big sensor distance, small rotate creates long, smooth curves like ocean currents.
- Sensor angle ≠ rotate angle destabilises the agents and the network constantly shifts.
- Drop agents to 1,000 to see individual paths; push to 30,000 for dense plasma.