Timothy Compston weighs up the critical elements that keep the wheels of dynamic pricing schemes turning in today's high-occupancy toll (HOT) lanes.
In the drive towards smarter tolling it is perhaps not surprising that sophisticated pricing algorithms are being rolled out to better reflect supply and demand on the roadway. This is the case with high-occupancy toll (HOT) lanes which a growing number of DoTs are seeing as a way of smoothing the operation of their existing, and planned, freeway infrastructure.
One leading exponent of dynamic pricing is
Regarding pricing, MnDoT’s acting regional transportation management centre engineer Brian Kary, says it has moved from relying on a vendor provided calculations to an in-house capability: “It was a bit of a black box before, not knowing how that algorithm was operating. We knew the general philosophy and what it was utilising but the finer details were somewhat unknown. Now our own software system is doing the calculations so it works similarly to the old system but we have more control as a DoT.”
Kary explains that the software focuses primarily on the volumes and speeds in the HOT lane rather than the general purpose lanes: “It updates the pricing every three minutes based on the worst location along that corridor, which is downstream of that pricing sign.”
Learning curve
John Hourdos of the University of Minnesota studies HOT lanes and pricing algorithms and is adamant that HOT lanes are the only realistic route to add additional capacity to today’s freeways: “You cannot just build new lanes because that capacity is soon consumed and you are back to where you started. The best way to introduce new capacity is through a managed lane where you control the demand and you allow it to enter and to develop slowly.”However, he says most HOT lanes development and congestion pricing happened in an ad hoc way: “In each case they design something. There are no guidelines; there are no tools. Each jurisdiction is doing its own thing. No one really knows if their method is good or if it is optimised.”
He has come to the view that fresh thinking is needed about the assumptions associated with existing models: “We have to develop things further based on real data, identifying actual driver behaviour and then trying to infer their thought processes when faced with that choice.”
When he compared the pre-implementation analysis for Minnesota’s HOT lanes with in-operation data, Hourdos found one somewhat counter-intuitive element stood out: “Up to a point we found that the higher the price, the higher the use.
The numbers seem to show that people were identifying the term ‘congestion pricing’ – which I find misleading in this case – and were basically thinking the price is higher and that means they are going to encounter worse congestion so they had better get in that [HOT] lane.” He says drivers do not understand the price is only based on the congestion in the HOT lane and takes no account of the congestion on the general purpose lanes.
I-85 lessons
When Georgia’s
Deciding how best to implement dynamic pricing was a steep learning curve for SRTA executive director and board secretary Christopher Tomlinson and his team: “Initially we wanted the [pricing] algorithm to look at traffic in the HOT lane and the adjacent general purpose lanes to have a predictive pricing strategy. Our fear was that if we only based it on traffic in the HOT lane it would be a lagging indicator of congestion.”
But according to Tomlinson that was a mistake: “We quickly changed the weighting to be primarily based on the level of congestion in the HOT lane itself.”
He explains that by taking both the HOT lane and general purpose lanes into account, the price set was higher than the level of congestion in the HOT lane really warranted: “Put another way, the system placed a value on the trip in the HOT lane based on the much heavier congestion in the general purpose lanes.”
Tomlinson recalls that during the first two or three days the price was in the region of US $5.35 yet the new HOT lane was seeing extremely light usage: “When we adjusted the parameters at the Governor’s behest, about three days in, the price fell down to about $1.55, in the same time period. When we opened the lane there were about 75,000 toll transponders but not all of them were going to be in the lane, in the corridor, at the same time. Relatively speaking that was a light amount of transponders so we did not have to price it as aggressively.”
On the new projects like the I-75 South Metro, Northwest Corridor and the I-85 Extension, Tomlinson says the SRTA is aims to apply pricing algorithms that look at data, almost exclusively from the HOT lanes: “Our weighting is 99% based on what’s in the HOT lane. Although this may be a lagging indicator, having seen how the prices change, it appears to balance out in the end,” concludes Tomlinson.
In California, the San Diego region is home to the I-15 Express Lanes which forms one element of a wider Integrated Corridor Management (ICM) system. Ray Traynor, director of operations at San Diego Association of Governments (SANDAG) says the principles applied to calculate the I-15 toll rates have remained reasonably constant: “Once we were able to apply electronic toll collection we have always had a toll per mile. This is dynamic with the amount charged based on the level of congestion. Basically the toll increases as congestion increases. The whole idea is that we are trying to maintain an average speed of at least 60 miles an hour [in the HOT lanes],” says Traynor.
Asked about whether the congestion focus is confined to the HOT lanes, Traynor says San Diego felt it made sense to consider what is happening in the general purpose lanes within the overall equation.
He stresses that travel time remains a key element and the savings drivers achieve by using the HOT lanes compared to the general purpose lanes: “That is really the whole idea of calculating a rate. We are attempting to assign a value to the travel time savings yielded from the express lane. The greater the time saving then the higher the fare.”
Beyond this, Traynor explains that tolling on I-15 is also geographically tailored so the rate drivers are charged will vary depending on where they join the lane: “The closer [the driver’s] entry point is to a congested segment, the greater their fare per mile will be.”
He stresses the importance of the rate being consistent, so the rate quoted to a driver prior to entry does not change during the duration of that journey whether the trip is for one segment or the entire length: “While the pricing is dynamic and changes, potentially every three minutes, we honour what is quoted.”
There have been some refinements to the I-15s tolling algorithm as a result of analysis conducted by the joint agency management team (which includes Caltrans). Traynor says this provided an opportunity to slightly extend the ‘step function’:
“This was about having a slightly higher density materialise in the [HOT] lanes before starting to raise the toll. What was happening was that when you get to a certain point, and congestion levels become so great, you effectively price single occupant paid users out of the facility.”
The roll-out of dynamic pricing is not without its issues but much more is now known about how algorithms can be applied and refined to keep prices at the right level to achieve the HOT lane objectives. What is still unclear is whether considering what is happening in adjacent lanes will help, or hinder, this process and the crucial importance of ensuring drivers feel they are being treated fairly as prices rise and fall.
Brian Scott, vice president for transportation at SRF Consulting, was SRF’s project manager on the I-394 HOT lanes project which was, essentially, a design/build type of contract. “We did everything from the concept to developing the access points. Our team members were involved in developing the algorithms. Three or four years later MnDOT went ahead with the 35W HOT lanes."
In 2005 he recalls that the HOT lane concept - and what was happening in Minnesota - was revolutionary and changed the thinking behind high-occupancy lanes. "As we were developing the I-394 lane, Washington State was looking at its north-south highway."
Michael Janson, senior transportation analyst at SRF, points out that the MnDOT lanes stand out as one of the first to be 'truly' dynamic: "I'm not sure if they were the first [HOT lanes] but in terms of pricing and the frequency of change, they were definitely one of the leaders in how often the price was adjusted."
Janson had focused on the challenges of HOT lane pricing implementation during his graduate student days: "Although I wasn't involved in the initial deployment [in Minnesota], I did research the pricing algorithm in graduate school. For MnDOT the algorithm just looks at the density in the HOT lane, it doesn't look at the densities within the general purpose lanes."
From his work, Janson stresses the importance of having good detection methods in place to deliver accurate traffic densities results: "You also need an ability to process that data quickly and then to display that adjustable price to drivers."
He adds that before dynamic pricing, previous systems relied on time-of-day or other pre-set schedules. "They didn't adjust on the fly."
Another pertinent issue highlighted by Janson is ensuring that the algorithm is not only accurate enough to fully utilise the HOT lane but to also smooth the price changes: "You don't want to have drivers using the HOT lane one minute for a dollar a mile and then 30 seconds later it is adjusted to five dollars or something like that," he explains. Ultimately, he believes, the recipe for successful HOT lanes is striking the right balance between performance, keeping the driving public happy, and making sure that there are not drastic changes to pricing along the way.
About the Author: Timothy Compston is a freelance journalist who writes on traffic technology and security issues.
