The New Mexico desert might not seem like a place where people would worry about the morning commute. There is no rush hour, there are no twisted freeway intersections and, in fact, there are not very many cars.
But scientists in New Mexico are finding traffic-jam solutions, without having to drive through crowded cities themselves. Instead, they and other traffic researchers are using clever math and speedy microprocessors to envision how people awash in turbulent seas of cars can better navigate the streets.
"Transport is in this funny area between physical and economic systems," said Christopher Barrett, a computer-simulation theorist at Los Alamos National Laboratory. "It is in a really cool place for serious scientific development."
Some of the mathematicians, computer scientists and physicists at the Los Alamos lab have shifted their talents to traffic because they decided it was time to address a national concern.
"We have major, major decisions to make related to transport," Barrett said.
There are large traffic-research teams in Germany, Canada and Scotland, as well as in the United States. In the next century, an increasing population, worsening pollution and diminishing natural resources will all contribute to countries' need to understand traffic and transportation, Barrett said.
To address these concerns, scientists are trying to solve the traffic problems in cities.
Once, untangling the web of traffic was just a matter of building more roads or replacing stop signs with traffic lights. But congestion in modern cities has grown so complex that those solutions aren't good enough any more.
Scientists have had to devise mathematical techniques and computer simulations to understand what drivers will experience in increasingly crowded road systems.
Clouds and traffic
Traffic jams can be predicted, for example, by estimating how many cars on a freeway will cause traffic to bog down.
To make this prediction, scientists borrowed the math used to describe water in a cloud.
In a cloud, gaseous water molecules travel freely from place to place, rarely bothering each other. But when many water molecules are packed into the cloud, they condense, forming drops of liquid water.
The process of water condensation, known in physicists' lingo as a "phase transition," is similar to the transition between free-flowing traffic and a jam.
The jams start to occur at a precarious point known as "critical density," where traffic patterns are hypersensitive to the number of cars on the road.
"If you add just one more person, it can lead to a tremendous increase in travel time," said Ismail Chabini, a professor in the Center for Transportation Studies at the Massachusetts Institute of Technology.
According to the Los Alamos model, that critical density of cars occurs when a road carries 20 percent to 30 percent of the maximum number of cars that could be put on it, said Brian Bush, a physicist working on the Los Alamos traffic project. The critical density is affected by the number of lanes, the speed of the cars, and erratic variations in driver behavior, he said.
Predicting critical density
Scientists' insights about critical density could help prevent jams from occurring. One solution to these problems would be to incorporate the new insights into "intelligent" transportation systems.
If those systems - which typically use underground car sensors, stoplights and other onramp barriers to control traffic flow - were programmed to more accurately calculate critical density, then they could stop just enough cars from getting on the freeway to prevent jams, Chabini said.
Another problem the scientists hope to do away with is the unpredictability of travel times. For example, on Monday, it might take 15 minutes to get to work, but on Tuesday it might take half an hour on the same route.
"There could be as much as a five-minute variability across a five-mile-by-five-mile area," said Dick Beckman, a statistician on the Los Alamos project.
But differences in road design could reduce that variability, Beckman said. Depending on the number of lanes on a freeway, for example, that varying 15-to-30-minute commute could be a steady 20 minutes.
Other traffic researchers are developing the mathematics to try to predict traffic patterns on existing roads an hour before they happen. That would be enough time for cars to change routes and avoid potential tie-ups, said Qi Yang, a research scientist at Massachusetts Institute of Technology.
Using computer simulations
For the Los Alamos researchers, the computer has been their laboratory of choice since they started the project four years ago. That's because it is difficult to count the comings and goings of vehicles out on the streets.
"Making any measurements on traffic is very expensive," Bush said.
Instead, the scientists create simulations of cars criss-crossing the grids of a city map. Their most recent undertaking has been a computer modeling of part of the Dallas-Fort Worth area. They described their work in two papers published in the past few months on the Los Alamos online research archive. One of the papers also will appear this year in the "International Journal of Modern Physics C."
The computer program they wrote for their project was fed simulated driving patterns of hundreds of thousands of people. The make-believe driving profiles were based on real survey data - including information such as what errands people ran, where they went to work, what time they went out and what routes they drove.
"It's a fictional Big Brother," Beckman said.
Driving virtual streets
The Los Alamos team's simulation is based on a framework of mathematics called cellular automata. The Los Alamos scientists were among the first to use this framework as a tool in very large traffic simulations.
Typically in cellular automata models, thousands of computerized drivers are programmed to drive virtual streets by following a set of rules dictated by the computer programmer.
For example, one of these drivers could be programmed to simply make a right turn whenever possible and to drive forward otherwise. That's the set of cellular automata rules for a robot that drives clockwise around a block. But a driver like that does not mirror real human behavior.
In the Los Alamos simulation, the rules are more complicated - and more realistic. The computer drivers are smart enough to speed up if the road is clear and to slow down if a car is in front of them. They also know how to change lanes, stop at traffic lights and use side streets on their way to a destination.
The Los Alamos scientists placed hundreds of thousands of these mock drivers in the computer-simulated Dallas area and let each one drive through the city. Then, all the scientists had to do was watch what happened.
"This is a different approach to modeling. The traffic characteristics emerge from the (individual) vehicles," Beckman said.
Future insights from this simulation would vary from city to city. But because it will be easy to add streets, traffic signals and freeways to the computerized map, city designers won't have to wait until construction finishes to see how well new road systems work.
"We are going to have a new set of analytical tools for regional planners," Barrett said.
The physics of traffic
While the cellular automata framework was borrowed from theoretical computer science, other traffic models were taken from physics, including viewing traffic jams as tidal waves in a stream.
In that view, streams of cars travel through roads the way water does in a city with canals. Heavy traffic creates a surge in the otherwise smooth waters of the canal. That can make the city's water - or cars - flow more slowly because of clogging pathways.
Proponents of these fluid models tout the speed with which computers can figure out how cars will swish through the city. That's because these models consider properties of the stream, not of individual cars, which makes their equations easier to solve.
Researchers hope that with the knowledge they're gaining, it won't be long before freeways are free for driving again.
