Variable Mesage Sign (VMS)

Destinations are determined based on dominant travel patterns. UDOT collects vehicle counts through radar sensors at freeway ramps and periodically along the freeway. UDOT also collects speeds statewide for every road classified as major collectors, arterials and freeways. These data sets are combined to identify dominant traffic patterns across the state. Dominant travel destinations are then considered for travel times based on how the destination can be represented on the sign. VMS are regulated by the FHWA Manual on Uniform Traffic Control Devices (MUTCD). The MUTCD has specific requirements for things like font size, color, messaging, spacing, abbreviations, etc. Those destinations that can meet these specific requirements are selected for the VMS Travel Times. Priority is also given to routes that provide reasonable alternatives, specifically at decision points along the freeway. These alternatives reach the same destination, are frequently used, and demonstrate lower travel times in comparison to the primary route several times over the past 6 months. 

UDOT's current traffic speed data, Cleardata, is provided through Iteris. This is a different data set from Google. Being different data sets, the algorithms to develop travel times are also different. However, different doesn't necessarily mean either one is wrong. Consider the following example: A blue car enters Southbound I-15 at 3300 S at 5:15 pm and is headed to Bangerter Highway. The predicted travel time on the VMS might be displayed as 30 minutes. Note that by the time the blue car gets to Bangerter Highway, the time is expected to be 5:45. This means that travel times are a prediction into the future.  A red car that enters at 5:25 pm, may experience a different travel time from the blue car.

Because travel times are a prediction of the future, there is no perfect way of doing that. We have a lot of data saying what it was yesterday or what the average travel time is for a normal day. But how do we know if today will be a normal day? And how much weight do we put into the historical data versus the data we are measuring now? Most data vendors, for this example, would consider the speed of each segment of roadway between 3300 S and Bangerter, all at exactly 5:15, and calculate an estimated travel time, assuming that those conditions are likely to stay the same. One key difference, though, is how much that travel time estimate gets influenced using historical averages. Another factor is how individual lane speeds are aggregated.  Along this example route, there tends to be a lane speed differential between the express lanes and the general purpose lanes. There also tends to be fewer cars in any drop lanes exiting the freeway, particularly at 5300 S, which can experience faster speeds during severe congestion because those vehicles are getting off of the freeway.  In general, UDOT has found that our current data set tends to be more conservative than Google. It uses harmonic mean speeds rather than arithmetic mean speeds which causes the travel times to react a bit more to speed discrepancies between lanes. We have found that Google seems to temper travel times a bit more both along the route and across lanes. UDOT is also able to provide more context to be used in our data set algorithms.  With speed and travel time data, there are always some assumptions that have to be made.  What if that blue car stopped for gas? Do we continue to include that travel time or not?  Because Google is a worldwide agency, they have to make some broad assumptions. We see this difference in our data more clearly along roadways that operate a bit differently from typical roads in the same class.  In the Cottonwood Canyons, for example, data vendors often make the assumption that a vehicle that hasn't moved for, lets say 20 minutes, must have left the roadway and therefore, their travel times should be excluded from the data set.  But we, here in Utah, in the winter, know that the little Subaru with its skis mounted on top, that hasn't moved for 20 minutes, is probably still there waiting, because they are headed for the ski resorts and that interchange exit is used mostly for ski resort traffic. So all of these things get considered a little differently and there isn't a perfect answer. And the algorithms used by various vendors are proprietary information so we can only evaluate the data results. That being said, UDOT periodically sends out pilot cars to measure travel times in the field.  All of our field observations fall within the same range as Google and our current data through Iteris.

Every 5 years, UDOT advertises a Request for Proposals to purchase traffic speed data statewide, for all major collectors, principal arterials and freeways from data curators. Data is typically collected from vehicle navigation systems and cell phone applications. UDOT Policy requires the data to be anonymous. Selection is based on data quality, data coverage and price. UDOT's current data provider is through Iteris and that data has proven to surpass data quality requirements.  At this time, Google has yet to offer the historical and real time data set that meets UDOT's requirements.

Travel times are generally measured from the physical Variable Message Sign to the gore of the ramp to the destination or to the actual gate or booth where applicable (example: Zion National Park). The gore is where the paint lines come together at a point between the ramp and mainline. In regards to how the travel times are actually calculated, consider the following example: A blue car enters Southbound I-15 at 3300 S at 5:15 pm and is headed to Bangerter Highway. The predicted travel time on the VMS might be displayed as 30 minutes. Note that by the time the blue car gets to Bangerter Highway, the time is expected to be 5:45. So travel times are a predication into the future and are therefore, an estimate. The travel time route is broken into segments along the roadway  The travel time is calculated using the speed of each segment of roadway between 3300 S and Bangerter, all at exactly 5:15, assuming those conditions are likely to stay the same. The travel time may also be adjusted if necessary, using historical average travel times for that time of day and that day of the week for that specific route. For this example, we know that the travel times at 5:15 on a Tuesday are far higher than the travel times at 5:15 on a Sunday. This becomes more important for routes that can see more dramatic changes in travel time between the time the blue car passes under the 3300 S VMS and when the blue car actually reaches the gore of the off ramp to Bangerter Highway.

UDOT evaluates the data quality of our VMS Travel Times every two weeks. We compare our data with other data vendors, like Google, to identify differences. We also compare our data with extrapolated travel times measured between radar sensors which provide us with point-speed data. This is more effective in areas with higher density radar availability. We compare closure and incident alerts developed with our travel time data to the closures, construction zones and incidents that are manually tracked by our operators at the Traffic Operations Center. Finally, we also compare our travel time data with our volume and speed data collected from our side fire radar and from our traffic signal detection to evaluate consistency.  When appropriate, differences are further researched and used to continue to tune our algorithms. These evaluations are currently being compiled into summarized reports that will be posted here when available.

Some routes in Utah experience dramatic drops in traffic demand during off-peak hours, to the point that the travel time messages do not provide nearly as much value. Unfortunately, lighting up the VMS also costs energy. In an effort to be energy conscious, and still provide valuable information, some of the VMS have been scheduled to run when they can provide the most amount of value for the cost to run them. You may have noticed that many of our VMS run for longer periods of the day. We have found that there is value in providing this information even during uncongested times to help set expectations for what travel times are during uncongested times. We are hoping this will help drivers plan ahead and adjust their schedule when possible to avoid delays when they can.  We do have some smaller VMS that are able to run on solar panels. But the energy to light up even these smaller signs is still significant. In an effort to preserve that energy, the solar powered VMS are not scheduled to run travel times. This way they are ready when they are really needed.

This is currently an issue related to the update of the ATMS software being used at the Traffic Operations Center and complications relating to font sizing and spacing. This is also why the travel time messages are not all configured to the old left and right justified format that we generally prefer. We are excited, though, that this new software allows us to separate travel times into hours and minutes rather than requiring everyone to do math while they are driving. Our old system would display, for example, 75 minutes. Now we can display that as 1 hour and 15 minutes. This has been a national trend that we are excited to be a part of. Having the messages displayed is a higher priority over specific formatting so we appreciate your patience as we work to fix these issues.

UDOT is always looking to do more with less. So finding additional uses for what we already have is a great place to be innovative. We are currently running a pilot to dynamically measure the speed of oncoming heavy trucks approaching curves to warn them if their speed exceeds the safe speed for that curve. UDOT is also currently testing new sensor technology that might allow VMS to be used to protect bridges from being hit by warning heavy trucks that they are too tall. These bridge hits are usually the result of equipment that has been loaded too high. Although the costs to repair those bridges are not fun to pay, the real driving force behind this effort is the potential safety improvements associated with it.