Sensory Perceptions
GDOT's Mark Demidovich talks to ITS International about the use of cellular probe-based technology to monitor traffic flows and detect incidents in real time
Cost versus capability is a perennial engineering headache. The trade-off of budgetary reality against absolute performance is nowhere more apparent than where the effective monitoring of long stretches of lightly used rural road is concerned; loop and more especially camera systems are simply too expensive to justify procurement in anything like the numbers which are needed to provide the density of coverage enjoyed in more highly populated urban areas.
In the US, under a project currently scheduled to run until the end of November, the Georgia Department of Transportation (GDOT) has been testing two cellular telephony probe-based solutions, one each provided by Airsage and Cellint, as possible infills between more traditional sensor systems in urban areas that can cost-effectively and anonymously gain better knowledge of what is occurring in near real-time on both strategic and non-strategic routes.
This is distinct from the more common use of probe vehicles to measure point-to-point travel times and very often - in Georgia's case - involves providing a continuous data gathering capability on routes where none has existed previously.
Geographical drivers: Georgia is the largest US state east of the Mississippi. It covers an area of some 59,411 square miles (154,077km2) and has terrain which ranges from mountains in the north, to flatlands, to coastal in the south-east.
Despite its size, it has only one major urban centre, Atlanta, the state capital, with a population of 5.2 million. Beyond that, there are only a few conurbations with populations that reach the 150,000-200,000 mark; the majority of the state is rural and sparsely populated. There are, however, about 1,250 miles (2,000 km) of Interstate and there are significant levels of freight and vacational traffic among the 100,000-150,000 vehicles per day which use the state's rural roads.
Away from urban centres, loops have historically provided the principal detection capability, says GDOT's Mark Demidovich, the man responsible for managing the cellular research project. Typically, these can be anything up to 50-75 miles (80-120 km) apart. The loops provide no real-time information; their main use is in gathering count data which GDOT is committed to do on behalf of the Federal government.
Probe-based solutions: The first flirtation with the idea of using real-time probe data gathered using cellular networks occurred about four years ago, when Atlanta-based Airsage approached GDOT with the concept.
Two years on from that, GDOT's research department decided to take the idea forward.
"My hypothesis was that if the technology could be shown to be both accurate and robust then it would completely change the detection game," says Demidovich. "There would be the opportunity to move away from roadside/embedded systems and use the vehicles themselves, by way of the cellphones being carried, to monitor what was going on."
Demidovich sees the technology as complementing and not replacing, loops and cameras. As ever, it is a question of application.
"We still need vehicle count information and while it may be possible to do something with the operating algorithms of the cellular solutions, loops and video detection remain a tried-and-true technology for us in that regard. Additionally, video and microwave continue to have their place for intersection detection associated with signal operation. The cellular solution can also be used for incident detection, however, as its ability to measure flow speeds can indicate that a problem is occurring."
Conducting trials:Testing involved the use of floating vehicle runs performed by an independent third-party consultant. Comparisons were made of cellular probe-based vehicle data with data obtained from Global Positioning System-equipped probe vehicles. Measurements were carried out in 10mph (16km/h) segments, to assess the systems' accuracy at different speeds and in order to address the concern that the cellphone-based solution would only really be effective at higher speeds.
Cellint is now stating that its TrafficSense solution was able to demonstrate an average speed difference of less than 5 per cent from GDOT's more conventional sensor systems in both congested and free-flow conditions; a report on the performance of the Airsage system will be published at the end of the project in November.
"We found that the percentage error varied according to speed category but overall accuracy was good," says Demidovich. "The average difference across all speed ranges was 5-7 mph [7-11 km/h] which, to me, is good enough to act as a feed to NaviGAtor, Georgia's real-time road information system; we allow a plus or minus 10 per cent figure when specifying video systems, for example, because there simply isn't a need for pinpoint accuracy."
Latency issues: Latency, says Demidovich, was one of his initial concerns. "We function in real time but that needn't mean second by second, especially when it comes to traveller information. We poll information from traditional systems every 20 seconds, update the trip time information on our variable message signs every minute, and refresh the website every three minutes.
"With the cellular solution, there are a few extra steps involved in getting information from the vehicle to the repository. The process takes about five minutes." (Cellint has claimed to have achieved a latency of just 60 seconds in another application of its technology in Israel.)
Costs and effectiveness: As well as being rapidly deployable in relative terms (see Sidebar, Speed of deployment), the costs of using the cellular systems have fallen dramatically over the last couple of years, according to Demidovich, from around US$10,000 to around $500 per mile. In terms of system saturation, he feels, there can never be 'too much' information.
"By contrast, too little might be an issue. If you've a stretch of rural highway being used at 3am by six cars, only two of which are carrying a mobile phone, you're going to get very little data out of the system. But with very little or nothing going on, do you then need an always-on monitoring system?" he concedes.
Further developments: As a follow-on to the current project, Demidovich is keen to push into more heavily built-up urban areas to explore the effects - if any - of canyoning and so forth.
"My concept is to push onto arterial routes because they're a very different beast to the Interstates. For example, arterial access is much less restricted - you have to contend with driveways - and there are intersections and traffic signals which introduce complete, utter and planned stops. I'm keen to see how this type of system copes in that context.
"Deployment is always going to be budget-dependent but I see major applications at the edges of urban areas - Atlanta is expanding all the time."
A particularly interesting scenario within the state, he notes, is the monitoring of the evacuation routes used in the case of natural emergencies such as hurricanes. This is a prime example of an instance where road use is very occasional and very heavy but when information is absolutely vital. It underlines the need for a monitoring capability on rural routes where the significant cost of more conventional solutions might be difficult to justify.
"It might also be dependent on the creation of new business models," says Demidovich, "as we'd be dependent on the willingness of the telephone companies and equipment suppliers to install systems on routes for which GDOT would perhaps not be willing to pay for constant flows of data. It might, though, if the suppliers find another way to sell on the data, still be a viable option."
Source: Original article (requires login). Published: Monday 15 October 2007
For more information please contact us.
