It is often claimed, by disgruntled programmers at least, that the problem with computers is that they only do exactly what you tell them. Loaded with enough rules, however, there are few tasks computers cannot overcome. The world’s fastest supercomputers are used to model problems as complex as climate change and nuclear explosions. But even some of their slower cousins are sufficient to simulate and forecast urban development patterns. And that is precisley the focus of civil engineering professor Dr. Eric Miller’s research.
Miller and his team are using an approach called micro-simulation to forecast the growth of cities. This is done by modeling the behaviour of thousands of individual agents in a large geographic region-think of it as a cross between video games SimCity and The Sims. “The aim is to create a tool for urban policy analysis allowing one to simulate and compare outcomes of different policy decisions,” says Miller. The strength of this approach is that it captures how a city’s transportation system and its urban form affect each other.
About 15,000 lines of C++ computer code define the set of rules governing the behaviour of these digital citizens, using concepts from psychology and microeconomics. Each agent behaves as an Homo economicus might, as they decide where to live, where to work, and how to get from point A to point B. Matt Roorda, one of Miller’s associates, is developing the part of the prototype that captures the way individuals in a household plan and coordinate out-of-home activities. “We’re trying to simulate how people behave, to predict the types of activities they are likely to engage in,” says Roorda.
To put their prototype to the test, Miller and Roorda developed a microsimulation of the GTA using 89,000 virtual households, extrapolated onto the entire city’s population, so as to estimate what travel patterns might look like in 2031. Assuming little or no changes in road and transit networks, the total number of car trips per day increase from 6.8 million in 1996 to 11.4 million in 2031, or from 72 percent to 74 percent of all trips. Meanwhile, public transit trips only increase from 1.3 million to 1.8 million per day; thus marking a decline in public transit’s overall share of future travel. Miller’s crystal ball, it would seem, points to a gridlocked future.
There are certain limitations inherent in such forecasts that must be kept in mind, however. They assume that underlying behavioural trends that have held true in the past will continue to do so-which is not always the case. Also, they cannot foresee random events, such as oil shocks or recessions, and their long-term implications. But if the program accurately simulates how people behave and make decisions, such hypothetical events can be introduced into the simulation, and changes in agents’ behaviour can be observed.
One interesting policy Miller’s urban simulator could test is road pricing schemes, such as the one recently proposed in Britain. A satellite tracking system would bill motorists for every mile, the toll depending on location, time of day, and distance traveled. The aim is to reduce gridlock nationwide by discouraging people from driving at peak times-paying up to £1.30 (C$3.07) per mile might deter one from driving when one could easily walk. The system won’t be in place before 2014, but Miller might find out beforehand if the tolls will actually work.
