Jeff's Website

The Future Of AutoCity

It became apparent that it is very difficult, if not impossible, to create a city with realism that can match the real world using a few simple parameters. The AutoCity program generates cities by using average values which are then deviated randomly to add realism. However in a real city, 'average values' are meaningless.

Consider, for example, a city such as London. We could specify an average building length and width by measuring every building in London and calculating an overall average. If we then generate buildings using this average size would we get something that resembles London ? The answer is, of course, no. An average is a mathematical figure. Buildings in the real world vary so much in their size that a calculated average can be meaningless.

A possible solution to this problem could be to create different 'types' of buildings (residential, office, school etc), and assign each type its own set of average values. These values could also be varied according to the buildings location in the city.

A more successful approach to creating a city model could be through the use of real-world data. Digital mapping data to quite a high resolution, as illustrated below, is available for many cities in the world.


If this data could be converted for use in AutoCity then there is no reason why highly accurate models of real cities, or areas of cities, could not be produced. Combine this with textures which could be photographs of the actual buildings themselves and we'd have a very realistic model.

The big advantage of AutoCity is that all the functionality required to produce this kind of model is already present. Everything within AutoCity is modelled from scratch using sets of coordinates, nothing has been pre-computed. Using the AutoCity API and writing filters to enable the import of some real world GIS data could produce very exciting results.