UPDATE: Thursday, May 13th, 2010
Travel time and speed estimation can be tricky, especially in light of bad weather, incidents, construction, etc. This is especially true in rural areas. For some time now we've been monitoring Google's "Traffic" layer and we've been reluctant to include it in OSS. It has included primarily urban information and, frankly, we have some serious concerns about its performance - especially if extended to rural areas. There now appears to be more rural coverage, so we've decided to enable this feature in OSS to informally evaluate how well it performs, side-by-side with the other data presented in OSS. (Realize that OSS is a research project, and this is an important topic for research.) At about the same time we turned it on in OSS today, we had a chance to compare it first-hand against real conditions - an incident near Corning, south of Redding. And, the results are mixed ...
Note the incidents shown via the Incident layer in OSS - we get this data from CHP:
The incident description for the Corning incident is listed as "BIG RIG LOST ENTIRE LOAD OF GLASS," and the incident is in the northbound lane.
Here we turn on the "Traffic" layer to show the Google Traffic data. Note several things. First, there isn't complete coverage of I-5 and there is no coverage of other roads in the area. Our understanding is that much, if not most, of the data is coming from cell phones running Google applications, and the density isn't high enough to make estimates for complete coverage. Note further that speeds on both sides of the incident are shown in "green", which indicates >50 MPH.
If we zoom in, we see an indication of reduced speeds - red and yellow segments in the northbound lane. Note that the southbound lane is showing no data - this is a bit difficult to see since the interstate is indicated in yellow.
Here we toggle the rest area layer to show proximity to the Corning rest area:
So - how did Google Traffic perform here? As stated at the beginning of this update, the results are mixed.
Ian Turnbull from Caltrans District 2 encountered this incident on his way to the office this morning - he phoned the incident in to the D2 TMC. Upon arriving at his office in Redding, Ian monitored the situation including the Google Maps Traffic display. The reduced speeds shown above appeared in Google Maps approximately 20-30 minutes after the incident, which is on the boundary between acceptable delay and information too late. As shown above, even when the delay appeared, the picture wasn't complete. The gap on the southbound lane is somewhat misleading. The biggest concern arose when the incident began to clear and traffic began to return to normal speeds. There was an hour or more delay before the clearance was reflected in the speeds and associated information.
What's the impact? A traveler seeing extended delay in this area may have chosen to divert to SR 99 via South Avenue through Corning, and then North to re-join I-5 at Red Bluff. This diversion would have cost 30 minutes, and may have been unnecessary.
Ian stresses what we refer to as the holy trinity of requirements for intelligent transportation systems - that the information must be accurate, timely and reliable. Does Google Traffic meet these requirements? Not yet, although it shows potential.
For now, we may leave the Google Traffic button on the site to evaluate further. But, if we observe more serious problems in achieving accuracy, timeliness and reliability, we may decide that it needs to be pulled.
Let us know your experience and opinion in regard to this. Feel free to email to: dgalarus "at" coe.montana.edu.