Forecasting traffic

There's a reason traffic and weather go together on virtually every local radio and television news program historical trends for both tell us what to expect in general, but actual current conditions can change quickly and unpredictably with unpleasant results if we don't keep an eye on them. Direct experience tells you whether you need a jacket or raincoat to go to work. Just stick your head out

There's a reason traffic and weather go together on virtually every local radio and television news program — historical trends for both tell us what to expect in general, but actual current conditions can change quickly and unpredictably with unpleasant results if we don't keep an eye on them.

Direct experience tells you whether you need a jacket or raincoat to go to work. Just stick your head out the bedroom window. And meteorologists have come a long way in accurately predicting near-term changes in the weather, so a quick check of the forecast will tell you if you'll need that umbrella later.

Traffic reporting, however, is still in the horse-and-buggy days compared to the high-tech tools we use to monitor, predict and disseminate weather information. Drivers can't see what conditions are like even a mile or two up the road. Using helicopters and a few video cameras mounted along the most congested roads are crude tools at best, most often alerting drivers to a problem when they're already caught in it or rerouting them around an incident that was cleared half an hour ago. And even if drivers get the information in time to act on it, there's no way to know if the alternate routes are any better or if they've already become choked with others trying to avoid the problem.

Getting wet because the weatherman blew a forecast may be uncomfortable, but usually you dry out with little or no ill effect. Having your trucks delayed because of congestion, however, can add significantly to the cost of doing business, not to mention damage customer relationships, hurt your fleet's reputation for reliability and push your drivers to look for a new line of work. And unfortunately, because so many of our highways and roads are now well above their design capacities, even the smallest hiccup in traffic flow causes a cascade of backups and roadblocks behind it, making congestion delays a routine part of almost every fleet operation.

The underlying causes of traffic congestion can only be addressed by concerted government and economic efforts to devise effective, long-term solutions that involve some tough decisions about infrastructure investment and road-capacity rationing. Realistically, any possible relief from those efforts is years, if not decades, away. That's not much help if you're losing 10%, 20%, 30% or even more of your fleet's capacity to increased travel times.

However, there are emerging information technologies that may help fleets recover some of that lost productivity, and a few are even making their debut as commercial services ready for integration with dispatch, routing and other operational management and planning tools.

Traffic, like weather, is about to become easier to plan for and even predict.


These new IT-based traffic services are best understood by dividing them into three categories. The first is real-time, which delivers traffic flow information based on actual conditions reported by multiple on-site sources. The second could be called short-term predictive, which looks out five minutes to an hour or so to inform drivers about expected changes in traffic flow based on those real-time reports. And the third is historical predictive, which uses collected historical data for specific road points during various points in the day, overlays it with scheduled events and offers detailed predictions on traffic flow to help fleets plan the most productive routes possible for any day or time of day. Each has potential value for commercial fleets, but different applications and different modes of data collection.

AirSage has been working on real-time traffic reporting for seven years, developing technology that collects anonymous speed and travel time information from cell phones out on the roads. An agreement with Sprint/Nextel currently gives it a large pool of cell phone users and it is currently pursuing similar agreements with other cell carriers in North America, Europe and Asia, according to Tom Bouwer, vp-sales and marketing.

“There are 220- to 230-million cellphone users in the country today, and some 50- to 60million of those phones are in cars moving along the highways most of the time,” he says.

The cell signals are fed to a central switch that might handle up to 2-million handsets. “A city like Atlanta might have two [switches] and Los Angeles 15,” says Bouwer. Those signals are then fed through AirSage's proprietary algorithms to create a real-time picture of both traffic flow and problem areas.

“The flow data gives you overall speed and travel times and incident data tells you about accidents and other problems,” says Bouwer.

Most commonly available incident data, whether it comes from a local transportation department or the local news station, will identify a problem, but provide no usable information on the severity, its impact on traffic on the particular roadway or if there's a viable alternative route. Flow data alone doesn't identify an accident or other roadway problem. Combine the two, though, with the right analytical system and you can look at any disruption that does occur, determine the severity and make a decision on changing a route, explains Bouwer.

While historical data on traffic congestion is useful for fleet route planning, “at best, it's only going to be accurate about 75% of the time,” says Bouwer. “The problem is that the other 25% — the accident or emergency situation — is exactly when drivers want information on how long the delay will be and if there's a better alternative. You need real-time information for that.”

Currently AirSage is selling its real-time traffic service to state departments of transportation, media outlets and in the very near future to service providers who will broadcast the data directly to onboard navigation systems.

“We provide it as an XML stream over the Internet, so we can sell it on a subscription basis to a wide range of application providers,” says Bouwer, who believes most fleets will receive their real-time traffic information as part of a package from a navigation or wireless data provider.


In the early days of developing “smart” or intelligent transportation system (ITS) highways, sensors embedded in the roadway were seen as the principal way of collecting traffic data. That approach quickly proved impractical, since coverage has been limited by the relatively high expense of installing and maintaining roadway sensors.

“Today sensors monitor less than 10,000 miles of our highway system, which covers over 150,000 miles,” says Bouwer. “Add in primary arterial roads — probably over 400,000 miles — and it's clear that sensors are becoming obsolete as traffic sensing technology for real-time information.”

Combining mobile wireless devices with the wireless network footprint of major cellular carriers makes it possible and practical to cover almost the entire roadway system at little cost. More importantly, the data collected by mobile devices is “computable,” making it easy to use for analytical and predictive systems.

“Traffic information from public sensors tends to be cognitive information intended for making announcements rather than computer bits of information,” says Alain L. Kornhauser, the founder of the mileage and routing provider ALK Technologies, as well as professor of operations research and financial engineering at Princeton University, director of Princeton's Interdepartmental Transportation Research Program.

“If a roadside sign tells you there's a stalled vehicle 22.5 mi. ahead, you don't know how good the alternatives will be because you don't know how long the delay is in most cases or what the travel time is on those alternative routes,” says Kornhauser. “Not only is the traffic data [from public sensors] sparse, but it isn't easily computable, so you can't use computers to do a comparison” in anything approaching real time.

Designed for smart phones, ALK's CoPilot Live combines GPS location data from the handsets with a hosted navigation application that provides turn-by-turn directions. Two years ago ALK enlisted 200 users for demonstration project that used smart phones as “probes” feeding in traffic flow information to a server.

“It was a focused group, all headed to the same destination,” says Kornhauser. “They were mostly concerned with (travel times on) three bridges crossing the Hudson River. With that probe information, it was a relatively simple computation to pick the right routes.”

The company is now getting ready to roll out the CoPilot Live, traffic feature, first in Europe, then in North America. “We're maintaining user experience and travel-time information for 1-million ‘segments’ that cover all major and minor arterial roads,” says Kornhauser. “I'd guess it's at least half-a-million miles.”

Initially, the real-time traffic information collected and analyzed by ALK's servers will be provide to all CoPilot Live users. “We see CoPilot consuming the information and providing the best routes rather than just giving you the data and making you choose,” the ALK founder says.

While the initial service will be focused on current traffic conditions, once the company builds a history for its one million segments, it can move to forecasting. “It's like weather forecasting, a completely parallel concept,” says Kornhauser. “Fifteen- to 60-minute forecasts should be right. I don't see a need for longer than that.”

While forecasting traffic based on current data is a complex problem, he points out that similar mathematical modeling is already used on Wall St. for stock forecasting. “With travel histories for segments and recurring patterns, this problem sets itself up well to do that,” Kornhauser says.

Because of the user numbers involved, Kornhauser sees both real-time traffic and forecasting developing first in the consumer markets. On the commercial fleet side, though, U.S. Xpress is already adopting CoPilot to provide its drivers with turn-by-turn navigation. “Traffic data is included in that roll out, so we'll see how they make use of it,” says Kornhauser. “In the future, you could even see carriers choosing to share that information” for more accurate truck-specific data.


Drawing on data from public sensors, toll tags and a network of GPS-equipped commercial fleets, Inrix is currently providing real-time traffic flow information for 94 U.S. markets and predictive information for 36 of those areas, according to Kush Parikh, the company's vp-business development.

Using intellectual property developed and patented by Microsoft, Inrix has built “a traffic fusion engine that creates historical data sets for 3.4-million miles [of roads] and overlays it with real-time flow data,” says Parikh. Launched in 2005, the Inrix system now has approximately 650,000 users in the U.S., including LTL and truckload carriers, service fleets and taxi companies that also feed it with GPS data.

While collecting traffic information from a wide base of mobile users may look easy, “the reality is that it's difficult to do because not all GPS data is created equally,” says Parikh. “Some may not be accurate enough, some may be collected on an archive basis. You can invest a lot of time and money in collecting [usable] information.”

Using its fusion engine, which it calls Smart Dust, Inrix is able to work with the various forms of GPS data currently being generated and then “fine tune it in our software,” Parikh explains. “That lets us [gather traffic data] at the scale we need without it being cost prohibitive.”

The real-time component of its traffic service is used by dispatchers to individually alert drivers of developing problems and perhaps route them around it or send them to an alternate location. While that can be valuable in handling unexpected problems, it's the predictive element that provides the most potential value for fleet operations.

Using the real-time GPS information, it's possible to predict traffic conditions out to about one hour, says Parikh. But the real value for most fleet operations comes from analyzing historical data to predict traffic patterns and flows far beyond that window.

“The vast majority of fleets are using our data for planning routes anywhere from a day or two to 30 days out,” says Parikh. Using the historical data sets it's built for its current 94 markets, Inrix then overlays predictive factors like weather forecasts, school calendars (the single largest influencer of traffic, according to Parikh), and local events such as baseball games or marathons to create long-range predictive models for route planning in 36 U.S. locations. “We have to build a separate model for each because the factors differ from city to city,” he says.

“A fleet may find that there are a few days every month when their standard routes don't make sense, so they can build better routes for those days,” Parikh explains. And irregular route operations can use the predictions to schedule off-duty times or deliveries to coincide with a region's peak traffic period.

Fleets then have the best of both worlds — predicted traffic conditions for preplanning the most effective routes and real-time data to alert drivers when the unexpected ties up traffic, says Parikh.

Although we still can't do anything about the weather, better information and forecasting has greatly improved our ability to deal with it.

Take your pick — real-time notice when there's a traffic problem, new routes based on the predicted impact of that problem, or forward-looking traffic data for planning more productive routes. Singly or in combination, these new systems have the potential to improve the productivity of any fleet on the road.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.