← Gallery

Slime Mould

Agents that leave a trail, follow the strongest trail they can see, and let the rest evaporate. Transport networks, with nothing planning them.

Output

A set shares every setting — paper, border, the sketch’s own knobs — and differs only in the seed. Regenerate any one of them without touching the others.

PNG: 3508 × 4961 px at 300 dpi. PDF: vector, RGB, 297 × 420 mm page, no bleed. Printed area 297 × 420 mm.

0 mm

Millimetres of unprinted paper on every side. The piece is composed to fit what is left, not shrunk to fit it.

Simulation

190

Cells across. Everything else is measured in these, so a finer grid is a finer network as well as a slower render.

9000
300

The network is coarse early and keeps refining. It never quite settles.

Agents

7.0

How far ahead an agent looks, in cells. The single biggest control on how coarse the network comes out.

22

Degrees between the centre sensor and the two side ones.

30

Degrees an agent turns when a side sensor wins. Below the sensor angle it takes several ticks to commit, which is what makes the veins smooth.

1.0

Cells per tick.

0.10

Chance per tick of turning at random instead of following the trail. Without a little, the network freezes into whatever it found first.

Trail

1.00

How much trail an agent leaves per tick.

0.200

Fraction of the trail lost per tick — what kills off the routes nobody is using. Turn it down and the network keeps merging until only three or four veins are left.

0.12

How much the trail bleeds into the four neighbouring cells each tick. Zero gives hairs rather than veins, because nothing an agent leaves is ever wide enough for another to find.

Drawing

The network is the trail map, not the agents — but watching a few agents walk it is the quickest way to see why it forms.

8
90

How many of the agents have their path recorded and drawn. All of them are simulated either way.

1.00
1.2

Colour

How it works

This is Jeff Jones’ 2010 model of Physarum polycephalum — a single-celled organism with no nervous system that nonetheless finds shortest paths through mazes. Each agent does three things. It deposits a fixed amount of trail into the cell it is standing on. It samples the trail at three points ahead of it, left, centre and right, and turns toward whichever is strongest. Then it steps forward. Between ticks the whole trail map blurs a little and fades.

It is stigmergy: the agents never communicate, they only alter the environment and react to it. The result is a transport network, because a route two agents happen to share gets twice the trail and so attracts a third. Diffusion is what makes that possible at all — without it a deposit is one cell wide and no other agent will ever find it — and decay is what stops it running away. Turn the decay down and the network keeps merging, and merging, until three or four enormous veins are all that is left; that behaviour is real, and the reason the default sits where it does.

One rule in the middle of it decides whether any of this works. When both side sensors beat the centre, the agent turns to a side at random. Turning to the stronger side instead sounds obviously better and is not: it funnels every agent within reach into whichever vein is brightest at that moment, and the sheet ends up with a handful of motorways and nothing else. What is drawn is the trail map rather than the agents — contoured, so the print file is a vector — but the paths of a few agents can be drawn over it, which is the quickest way to see why the veins are where they are.

R regenerate · S PNG · P PDF