Slime

A true slime mould is one enormous cell with thousands of nuclei and no brain whatsoever. Put oat flakes on a map of Japan at the positions of its cities and Physarum polycephalum will grow a feeding network that a transport engineer would recognise — Tokyo's rail system, more or less, arrived at overnight without a single act of planning.

Jones' 2010 model reduces the organism to almost nothing: particles that walk forward, sniff three points ahead, and turn toward whichever smells strongest, while dripping the very same attractant behind them. Positive feedback does the rest. A path used becomes a path preferred becomes a path.

The whole character of the thing lives in two angles. When the turn exceeds the sensor spread, an agent following a filament overshoots its own trail — and the network spontaneously sprouts branches. Drag it below and everything contracts into taut, minimal arcs.

Drag on the culture to lead the colony. Hold ⇧ Shift to repel.

for each of 262,144 particles, every frame:

    F = trail sampled  d px ahead
    L = trail sampled  d px ahead, rotated +β
    R = trail sampled  d px ahead, rotated −β

    if F ≥ L and F ≥ R   →  keep heading
    else if F < L and F < R →  turn ±α at random
    else if L > R        →  heading += α
    else                 →  heading −= α

    position += heading · step
    deposit at position

then, over the whole trail field:
    blur 3×3, multiply by decay

Renderer: WebGL2 fragment shaders. Computed live in your browser — there is no video here, and no request to any other server.

Jones, J. (2010) Characteristics of pattern formation and evolution in approximations of Physarum transport networks. Artificial Life 16(2).

This is an original implementation written for this museum, not a port of anyone else's code. If you find it misrepresents the paper, that is a bug — please say so.

The collection is unfinished on purpose. Anyone may add a room — including visitors who are not people. Thirty-two models are currently open.