A re-examination of the NetLogo Termites model

From Santa Fe Institute Events Wiki

CSSS 2006 Santa Fe

"A Re-examination of the NetLogo Termites model"

Doug Hoskins, The Boeing Company

The NetLogo "Termites" model project {in Models Library | Biology } is described on the information tab as:

This project is inspired by the behavior of termites gathering wood chips into piles. The termites follow a set of simple rules. Each termite starts wandering randomly. If it bumps into a wood chip, it picks the chip up, and continues to wander randomly. When it bumps into another wood chip, it finds a nearby empty space and puts its wood chip down. With these simple rules, the wood chips eventually end up in a single pile.

When running this simulation at slower than normal speed, we happened to notice that the generally moved in the random walk described, but would occasionally move with a sudden "burst" of speed in a straight line. This lead to a re-examination of the source code, and the identification of a number of model parameters that are not discussed in the model.

This project will consist of an analysis of the sensitivity of the stated qualitative results ("all the wood chips end up in one pile") to these undocumented parameters and search for an analytical explanation of the observed results.

To date, we have identified four (undocumented) features that seem to actually control the rate and characteristics of the "wood-chip collection behavior" being simulated. These are:

1) Ignore-own-pile-once After dropping off a wood-chip, a "termite" must encounter at least one empty space before it picks up another wood chip. This additional state in the turtle's process (state-transition-diagram) has the effect of making it impossible for a turtle to take a chip from the pile where it has just dropped a chip (unless its random walk brings it back to the pile). This behavior is hard-coded into the original simulation, and has been converted to a uer switchable boolean

2) Post-pick-up-run After picking up a chip, a turtle will select a random direction and run this distance before beginning it's search for a place to deposit its wood chip. The hard-coded value for this is 20 distance units.

3) Post-drop-off-run After dropping off a chip, a turtle will select a random direction and run this distance before beginning it's search for a new chip to pick-up. The hard-coded value for this is 20 distance units.

4) Wandering in the pile After finding a chip and encountering a pile, the turtle must then wander in steps of unit length until it finds an empty patch to place the chip. In the limit, as piles become large, this means the turtle has a 50% probability of beginning a random walk inside of the pile, searching for an edge. This random walk reinforces the circular shape of the pile and can entrap a significant fraction of the termites as the pile becomes large.

One other note is that termite piles tend to be [U]piles[\U] -- they are higher in the middle than at the edges. [time allowing] we intend to extend the model to allow multiple "chips" on a single patch, and probabilistic transitions between picking up and dropping wood chips. We expect this to generate the advertised qualitative behavior with a set of rules which is simpler, when all the rules are considered.

A powerpoint file showing the differences in behavior is here: [[1]]

The same features exist in the current termite sample project on the MIT media lab's starlogo page [[2]]