Thursday, May 01, 2008

Calculation Debates...driving edition

I was reminded on my drive to Chicago today of a fascinating road feature that always gets my mind stirring-- those road signs that tell you how long it is expected to take to get from where you currently are to a well-known place of interest. They are scattered throughout the Dan Ryan Expressway here, and they are used in the San Francisco Bay Area as well. What amazes me is how accurate they are.

Think of all of the variables that can affect how long it takes you to get from one place to the next-- and that they are constantly changing-- and to me, it's a small miracle that they are rarely more than a minute off one way or the other. I'm convinced this can't be the doing of any municipality; it's too useful and too effective. I bet communists love these things, too-- if we can calculate the complexities of rush hour traffic with great precision, surely an accurate Input/Output model is on the horizon, right?

Does anyone have any insight into the guts of these things? Is it any more than measuring the speed of passing cars at a few points along the road?


Justin M Ross said...

The distance signs on I-79 southbound between Pittsburgh and Morgantown become wildly inaccurate. You are actually much closer to Morgantown than the sign tells you you are.

Anonymous said...

They are called Dynamic Message Signs. An excerpt is provided below. For the full scoop (on Chicago -- I imagine the others are done in a similar fashion), see the following link:

Three sources of data are currently used by ISTHA to provide travel time estimates:

I-PASS readers at toll plazas.
Data from the IDOT loop detector network - to generate travel time estimates for non-Tollway roads in downtown Chicago.
The travel time system organizes the toll network into roadway segments, which represent the finest resolution for which data will be provided; roadway segments are typically bounded by ramps and mainline plazas. Lengths of segments vary across the Tollway (shorter near Chicago, but longer in outlying areas). The IPASS and RTMS data provide a level of redundancy for along certain segments of the toll system (i.e., there may be two sources of travel data for a particular section). ISTHA's travel time software allows the TIMS staff to choose which data source should be used in a given situation. Decisions are made by TIMS staff based on operational experience and judgment. For other segments, only one source of speed data is available (i.e., a segment may have only an RTMS station).

Segment-specific travel time data is then analyzed by ISTHA's travel time software at the TIMS Center to generate a travel time to a target destination from each DMS3. A travel time estimate to a target destination might include segments for which travel time has been determined using any one of the three sources of data. Travel time updates are currently posted to the DMS every 3 minutes, with the goal being to eventually post updates every minute. Figure 4 shows the travel time posting screen an operator views in the TIMS software.

Accuracy of travel time estimates has been a concern for ISTHA from the beginning of the project. During the early stages of system deployment, video observations of traffic flow were compared with the TIMS software's travel time estimates to provide reasonable assurances that the information was accurate. ISTHA also directed their on-call consultant to conduct a probe-vehicle travel time test to measure the accuracy of the travel times provided by their algorithm. The results of this test indicated that the travel times produced by the system were sufficiently accurate for release to the public. System operators conduct no ongoing reviews of data quality unless obvious discrepancies occur. However, data quality is checked annually using probe vehicles

Bob said...

I don't think it's that complicated to predict the time it will take to go from A to B once someone's already at A and the traffic guys (or their computers) can see what lies between A and B at that time.

The real trick would be to predict accurately the traffic times from A to B at 4:47 p.m. a week from today. No that I'd like to see.