Traffic (18 page)

At some point you may have come to a highway on-ramp, expecting to join the flow of traffic, only to be stopped by a red light. Such devices are called ramp meters, and they are found from Los Angeles to South Africa to Sydney, Australia.

Ramp meters often seem frustrating because the traffic on the highway appears to be moving just fine. “People ask me, ‘How come you’re stopping me at the ramp meter? The freeway is free-flowing,’” says Dawn Helou, the Caltrans engineer. “The freeway is free-flowing because you’re stopping.”

This is one of the most basic, and often overlooked, facts about traffic: That which is best for an individual’s interest may not be best for the common good. The game traffic engineers play to fight congestion involves fine-tuning this balance between what is “user optimal” and what is “system optimal.” This happens on several different levels, both having to do with congestion: how traffic moves on roads and how larger traffic networks behave (an idea I’ll return to in a later chapter).

The reason why highway ramp meters work is, on the face of it, simple once one knows a few basic facts about traffic flow. Engineers have been trying to understand, and model, traffic flow for many decades, but it is a huge and surprisingly wily beast. “Some puzzles remain unsolved,” declares Carlos Daganzo, an engineer at the University of California, Berkeley. The first efforts merely tried to model the process known as “car following.” This is based on the simple fact that the way you drive is affected by whether or not someone is in front of you, and how far away or close they are. Like ants responding to the presence of pheromones on the trail, you’re influenced by the driver ahead, a constant, unsteady wavering between trying not to get too close and trying not to slip too far back. Now imagine those interactions, plus lane shifts and all the other driving maneuvers, a fluctuating mix of vehicle speeds and sizes, a wide range of driver styles and agendas, a dizzying spectrum of differing lighting and weather and road conditions; then multiply all this by the thousands, and you can begin to appreciate the higher-order complexities of traffic modeling.

Even the most sophisticated models do not fully account for human weirdness and all the “noise” and “scatter” in the system. Traffic engineers will offer caveats, like the disclaimer I saw at one traffic conference: “This model does not account for the heterogeneity of driver behavior.” Do you feel uncomfortable driving next to someone else, and therefore speed up or slow down? Are you sometimes willing, for no apparent reason, to ride quite close to the car in front, before gradually drifting back? All kinds of strange phenomena lie outside easy capture by the traffic sensors. Car following, for instance, is filled with little quirks. A study that looked into how closely passenger-car drivers followed SUVs found that car drivers, contrary to what they said they did—and despite the fact that the SUV was blocking their view of the traffic ahead—actually drove closer to SUVs than when they followed passenger cars.

Or take what Daganzo has called the Los Gatos effect, after an uphill stretch of highway in California. You may have experienced this: Drivers seem reluctant to abandon the passing lane and join the lane of trucks chugging uphill, even when they are being pressured by other drivers, and even when the other lane is not crowded. What’s going on? Drivers may not want to give up the fast lane for fear of having trouble returning to it. They may also be unsure whether the person behind truly wants to go faster or is just keeping a tight space to prevent someone else from passing. A tight “platoon” forms, but for how long? We all see these odd patterns. One of the idiosyncrasies I have noticed in traffic flow is something I call “passive-aggressive passing.” You’re in the passing lane when suddenly the driver behind you pressures you to move into the slower right-hand lane. After you have done so, they then move into your lane, in front of you, and slow down, thus forcing
you
to pass
them.

The basic parameters of how highways perform have been gradually hammered out. One of the key performance measures is volume, also called flow, or the number of vehicles that pass a buried sensor or some other fixed point on the highway. At four a.m., before rush hour, cars may be zipping along a highway at 75 miles per hour. The volume is measured at 1,700 cars moving past a point in one hour. As rush hour begins, the volume quite naturally begins to rise in an upward curve, reaching a theoretical maximum of 2,400 cars traveling at 55 miles per hour. System-wise, this is traffic nirvana. Then, as additional vehicles enter the highway, the curve begins to drop. Suddenly, the volume is back at 1,700. This time the cars are going 35 miles per hour. “So you have the two 1,700s,” Helou says. “Same volume, completely different situation.”

Because traffic moves in time and space, measurements like volume can be deceiving, as can the highway itself. Solo drivers sitting in a highly congested lane may look to the HOV lane next to them and think that it’s empty—a psychological condition so prevalent it even has a name, “empty lane syndrome.” Many times it just seems empty because of the large headways between vehicles moving at much higher speeds. That lane may actually be achieving the same volume as the lane you are in, but the fact that the drivers might be going upward of 50 miles per hour faster creates an illusion that it’s being underused. Of course, neither of these positive or negative individual outcomes—the driver whisking along at 80 miles per hour or the people stuck at 20 miles per hour in the congested lanes—are what’s best for the entire system. The ideal highway will move the most cars, most efficiently, at a speed just about halfway.

Even as rush hour kicks in and the speed-flow curve begins to drop, traffic can perk along at what has been called “synchronized flow,” heavy but steady. But as more vehicles pile onto the highway from on-ramps, the “density,” or the number of cars actually found in a one-mile stretch (as opposed to passing a single spot), begins to thicken. At a certain point, the critical density (the moment, you will recall from before, when the locusts began their coordinated march), the flow begins to break down. Bottlenecks, fixed or moving, squeeze the flow like a narrowing pipe. There are simply too many cars for the road’s capacity.

Ramp metering aims to keep the highway’s “main-line flow” below the critical density by not letting the system be flooded with incoming on-ramp cars. “If you allow unimpeded access, then you have a platoon of vehicles that are entering the main line,” says Helou. This means not only more cars but more cars jockeying to merge. Studies have shown that this is neither predictable nor always cooperative. “That [merging] eventually breaks down the right lane,” she says. “This overflows to the next lane, because people try to merge left before they get to it. And then the people in the second lane try to merge to the next lane before they get to it, so you break down the whole freeway.” A line of cars waiting to exit an off-ramp can trigger this same chain reaction, one study showed, even when all the other lanes were flowing nowhere near critical density.

If done properly, ramp metering, by keeping the system below the critical density, finds that sweet spot in which the most vehicles can move at the highest speed through a section of highway. Engineers call this “throughput maximization.”

A simple way to see this in action involves rice. Take a liter of rice and pour it, all at once, through a funnel and into an empty beaker. Note how long it takes. Next, take the same rice and pour it not all at once but in a smooth, controlled flow, and time that process. Which liter of rice gets through more quickly? In a demonstration of this simple experiment by the Washington DOT, it took forty seconds for one liter of rice to pass through the funnel using the first method. The second method took twenty-seven seconds, nearly one-third less time. What seemed slower was actually faster.

Rice has more to do with traffic than you might think. Many people use water analogies when talking about traffic, because it’s a great way to describe concepts like volume and capacity. One example, used by Benjamin Coifman, an engineering professor at Ohio State University who specializes in traffic, is to think of a bucket of water with an inch-wide hole in the bottom. If the inflow into the bucket is half an inch in diameter, no water will accumulate. Raise it to two inches, however, and the water rises, even though some water is still exiting. Whether we drive into a jam (or a jam drives into us) depends on whether the “water”—that is, the traffic trying to flow through a bottleneck—is draining or rising. “As a driver, the first thing you encounter is the end of the queue,” Coifman told me. “The first thing you encounter is wherever the water level happens to be that day.” The bucket metaphor also teaches us something else about traffic: No matter how much capacity there is in the rest of the bucket (or on the roads), the size of the hole (or the bottleneck) dictates what gets through.

At places like bottlenecks, however, traffic acts less like water (it does not speed up as highway “channels” narrow, for one) and more like rice: Cars, like grains, are discrete objects that act in peculiar ways. Rice is what’s called a “granular media,” a solid that can act like a liquid. Sidney Nagel, a physicist at the University of Chicago and an expert in granular materials, uses the analogy of adding a bit of sugar to a spoon. Pour too much, and the pile collapses. The sugar flows like a liquid as it collapses, but it’s really a group of interacting objects that do not easily interact. “They do not attract one another,” says Nagel. “All they can do is scatter off one another.” Put a bunch of granular materials together, and it is not easy to predict how they will interact. This is why grain silos are the building type most prone to collapse, and it’s also why my box of Cascadian Farm Purely O’s cereal begins to bow outward at the bottom after several pours.

Why does the rice jam up as you pour it into the funnel? The inflow of rice exceeds the capacity of the funnel opening. The system gets denser and denser. Particles spend more time touching one another. More rice touches more rice. The rice gets “hung up” from the friction of the funnel walls. Sound familiar? “That’s like cars on the highway,” says Nagel. “And when you get narrowing of traffic, then that becomes very much stuff trying to flow through the hopper.”

Pouring less rice at a time—or moving fewer cars—keeps more space, and fewer interactions, between the grains. Things flow faster. As intuitive as the “slower is faster” idea is, it’s not always easy for a driver stuck in traffic to accept. In 1999, a state senator from Minnesota, claiming that ramp metering in the Twin Cities was doing more harm than good, launched a “Freedom to Drive” proposal that called for, among other things, shutting down the meters. The legislation died, but under another bill a ramp-meter “holiday” was declared. For two months the meters were turned off. Drivers could enter the highway at will, on so-called sane lanes, unfettered by troublesome red lights. And what happened? The system got worse. Speeds dropped, travel times went up. One study showed that certain highway sections had double the productivity with ramp meters than without. The meters went back on.

The “slower is faster” idea shows up often in traffic. The classic example concerns roundabouts. Many people are under the mistaken impression that roundabouts cause congestion. But a properly designed roundabout can reduce delays by up to 65 percent over an intersection with traffic signals or stop signs. Sure, an individual driver who has a green light may fly through a signalized intersection much more quickly than through a roundabout. Roughly half the time, however, the light will not be green; and even if it is green there is often a rolling queue of vehicles just starting up from the previous red. Add to this such complications as left-turn arrows, which prevent the majority of drivers from moving, not to mention the “clearance phase,” that capacity-deadening moment when
all
lights must be red, to make sure everyone has cleared the intersection. Drivers do have to slow down as they approach a roundabout, but under typical traffic conditions they rarely have to stop.

In the 1960s, experiments were made at the Holland Tunnel, one of the main arteries for traffic coming into and leaving New York City. When cars were allowed to enter the tunnel in the usual way, with no restrictions, the two-lane tunnel could handle 1,176 cars per hour, at an optimal speed of 19 miles per hour. But in a trial, the tunnel authorities capped the number of cars that could enter the tunnel every two minutes to 44. If that many cars got in before two minutes were up, a police officer made the next group of cars wait ten seconds at the tunnel entrance. The result? The tunnel now handled 1,320 vehicles per hour. (I will explain why shortly.)

On streets with traffic signals, engineers set progressions with a certain speed in mind that will enable the driver to hit a line of constant greens. To drive faster than this only ensures that the driver will be forced to come to a stop at the next red light. Each stop requires deceleration and, more important, acceleration, which costs the driver in time and fuel. A queue of drivers stopped at a light is a gathering of “start-up lost time,” as engineers call it (in an appropriately forlorn echo of Proust). The first cars in a queue squander an average of two seconds each, two seconds that would not have been lost had the car sailed through at the “saturation-flow” rate. The first driver at a light that turns from red to green, because he must react to the change, make sure that the intersection is empty, and accelerate from a standstill, generates the most “lost time.” The light is green, but for a moment the intersection is empty. The second driver creates a bit less lost time, the third driver less still, and so on (assuming everyone is reacting as soon as they can, which is not a given). SUVs, because they are longer (on average, 14 percent longer than cars), and take longer to accelerate, can create up to 20 percent more lost time.