The Settlement Simulation is a computational model based on multi-state cellular automata. The model uses simple behavioral rules to recreate the aggregation logic of dwellings and small subsistence farms in a given field via a self-organizing ‘vernacular’ methodology. The goal of the simulation is to investigate the sorts of distributions and collective form that might result while still meeting the fruit and vegetable requirements of the population. We work with fruits and vegetables for the time being, as historically they are hardest to transport and benefit most from urban environments (Steel 2009). Grains and livestock are more suitable to larger, peri-urban plots and require additional processing before consumption. Strategies for these will be explored in future development.

Settlement Simulation Rules.

Computational Rules

The simulation begins with a ‘settler’ arriving on a 1km2 site, placing their dwelling, and farming the land around it. The dwelling has a footprint of 72 m2, based on a typical urban townhouse and household size of 2.59 persons based on New York City’s average (US Census Bureau 2000). The baseline ‘productive’ area is based on a conservative 100 m2/person (Viljoen 2005).

A second settler arrives and ‘randomly’ chooses to either build adjacent or remote. Should they build adjacent, they need to relocate any farm land displaced by their dwelling. If remote, they will begin a new settlement cluster. This process is repeated until the target population density is met for the given field. Dwellings and farms are broken into modular units with each dwelling equal to one cell. With every one dwelling placed, three productive units are placed.

Friendliness Factor

The likelihood for a new settler to build remote or adjacent is weighted by a ‘Friendliness Factor’ (F), a number between 0 and 1. F=0 means all units will try to build remote, while F=1 means all units will try to be attached. An F of 0.6 results in a 60% chance that a unit will choose to build adjacent. Vertical growth is triggered when a settler ‘chooses’ to build on an existing dwelling in a settlement cluster which is has surpassed the user specified local density threshold.

F=0 (left) results in a homogenous distribution, while F=1(right) generates a single dense dwelling cluster surrounded by farms.

Density and Production Intensity

In order to achieve higher population densities, Production Intensity of a given plot must be increased. Higher productivity/unit area can be achieved in a number of ways, from innovative co-planting techniques (low energy/investment) to multi-story artificially lit/heated aqua/aeroponic greenhouses (high energy/investment) and a variety of methods in between. The system conserves energy by keeping production intensity minimal, stepping it up only as needed.

When the field is full, the algorithm allows agents to increase the production intensity of existing productive cells. Therefore, a settler’s productive needs must be met by either: an increase of yields from an existing plot, or a neighboring dwelling must incorporate production, introducing ‘hybrid’ types. Production Intensity (PI) is allowed to increase in steps of one, corresponding to number of people whose annual fruit and vegetable needs are met per 100m2. Therefore a cell having PI =3 means three people are provided for on that 100m2 plot.

Tissues generated with three Density(D) and three Friendliness(F) settings, showing the effects on local clustering and continuous farm sizes.


With producer and consumer cells distributed, the resulting pattern is analyzed to inform a network topology which will facilitate the movement of food with minimal energy expenditure.

Two distribution node types are generated: Wholesale Nodes (on-site farm shops) and Retail Nodes (small grocers located throughout the dwelling clusters).

Nodes are dispersed so as to minimize travel distances and encourage walking and biking. Wholesale nodes are located on large continuous productive areas (min. 1-2 hectares). Retail nodes are distributed throughout dwelling clusters such that any dwelling should be within a one minute walk, or 83m at 5km/hr.

Wholesale(W) and Retail(R) nodes are located based on distributions of dwellings and farms.

In order to connect the retailers back to their sources, a branching network is generated with the wholesale node at the root and retail nodes at the branching junctions. The angle between the branches controls the amount of detour in the network, and allows the designer to weigh energy expenditure resulting from inversely related metrics of Average Path (Trip) Length (consumer trips) versus Total Network length (infrastructure expenses and maintenance).

A larger detour (branching) angle results in longer individual trips but a shorter overall network.

The resulting network structures are ‘trees’, emphasizing ‘producer to consumer’ routes. However a healthy city must facilitate the social movement of people as well, therefore an Overlap Range (Ov) is implemented which connects terminal nodes of trees to adjacent trees within the specified range, introducing continuity and circuitry (the possibility for loops) into the urban circulatory network.

Individual distribution trees are joined to nearest neighbors to facilitate social movement.