Altering the Initial Placement of Agents

• Source Code for Location Model Same Income (542 KB)
• Shapefile (18 KB)

In many exploratory agent-based models, agents are initially randomly located in an area (e.g. Kii and Doi, 2005). This section explores the sensitivity of the model with regard to initial placement of agents in terms of specifically how the pattern of land-use changes by altering the initial location of agents. The same area and model is used as in the previous simulations (see the Effects of Space and Income) with the same number and type of agents, although the initial distribution of agents throughout the area is different. Specifically, the two most accessible areas (i.e. centre rings) are empty while the two outer rings are occupied. Table 1 highlights the altered initial starting conditions of the model (see Figure 1 in the Effects of Space and Income for a comparison of agent distribution). The parameters within the model were kept the same as with the previous simulation runs with 75m neighbourhoods and space requirements were set to the original values (see Table 1 from the Effects of Space and Income page) allowing for direct comparison with the previous simulations.

Table 1: Initial starting conditions with the two centre rings empty.

Table 2 compares the average distribution of agents after 100 iterations where the initial states of the centre rings or the outer rings were empty (see the animations below for how these patterns evolved). Overall there is very little difference by 100 iterations in the distribution of agents. In both cases, the outer four rings become empty over time (e.g. see the animations below which captures this change for 550 iterations). This is a result of agents’ movement and their desire to live in the most accessible place they can afford where their preferences for space are satisfied throughout the course of the simulation.

Table 2: Average distribution of agents at 100 iterations with different configurations of agents.

This section has explored how from a random scattering of agents, a structure appears similar to that predicted in the Alonso bid-rent theory. Employers can out bid residents for land with residents forced to more peripheral locations, and agents with greater space requirements are forced to move to more peripheral locations when income ranges are the same. The initial distribution of agents was explored and it appears there is little effect on the final spatial distribution.

A selection of animations up to 100 iterations where the two centre zones are left empty.

[TOP]

A selection animations up to 550 iterations: note how the distribution of agents changes little between 100 and 550 iterations.

Download zip file of images from model runs.

[TOP]

References

Kii, M. and Doi, K. (2005), 'Multiagent Land-Use and Transport Model for Policy Evaluation of a Compact City', Environment and Planning B: Planning and Design, 32(4): 485-504.

[TOP]