Dear John,

I am trying to compare travel time/speed for my calibration. Even though the model generates appropriate turn volumes, the travel time and speed are way off from the field collected data. Can you suggest any changes i can do to replicate the field conditions.

Hi Ravi,

It is difficult to say where the differences lie. There are a wide variety of reasons, of course one being when data was collected and when travel runs were gathered. If inputs such as volumes and timings were collected on one day, and travel runs on another, it may not be fair to compare.

Also, you might want to use the arterial measures of effectiveness. The problem is in how travel time results are gather. It will differ between version 5 and 6. In V5, the travel time along an arterial was simply the sum of through travel times at each intersection in the route. In the model, these throughs included vehicles coming from the upstream through, left and right. When you drive this in the street, you probably are always just going through to through. Therefore, I would expect SimTraffic 5 results to show a longer travel time.

In V6, this has changed slightly to reduce this effect. More details can be found in the Help file in the section on Reports, Graphics and Moes, topic on Measures of Effectiveness.

Thanks,

John.

Hai John,

Thanks for the reply. And i do agree the travel times will change from the day data collected and the day volumes were collected.

I am using version 6.0. The travel time from the field are showing around 35mph while the model speeds are 11mph.

Please let me know which parameters i can play with to get travel time and speed. I do tried some combinations of input parameters but i did not find any change.

Thanks,
Ravi.

Hi Ravi,

Wow, that is a big difference. Most calibration changes would probably not yield a difference big enough to show a difference. So, the differnce could be based on volumes, flows and/or timings.

1) Are the volumes in the model realistic when compared to the day the travel runs were performed? Maybe the volumes during the travel run day were smaller due to some unknown upstream meter.

2) Maybe the drivers are traveling more aggressively then the model is predicting. You could try reducing the Headway Factors in Synchro which will increase the flow in SimTraffic.

3) If the offsets in the model are off, this could have a large impact on travel times. Also, other timing parameters could impact travel times.

4) Finally, all simulation models are just that, models. Mathematical algorithms are applied to attempt to model driver behavior. It is possible that some unique situation exists in the model that is causing this larger delay at some intersection then normal. Do you see any spot locations stand out that are causing this problem? This can usually be visualized with an animation. There may be a work-around solution for cases like this. Feel free to send us this file if you think this is the problem.

Thanks,

John.

I am curious.
- What is the posted speed limit in this case? For most of the drivers along the road, that is the minimum to average speed, but it might not be same for synchro/simtraffic, i guess.
- The traffic counts and travel time runs should be made at the same time. were they? were they done on same season?
- I would look the field peak hours from the traffic counts also.
- How many runs were made for the travel time runs. I would check the standard deviation also.
I would exclude any outlier.
- Is there any left turns from onelane to a driveway or unsignalized road from the corridor? for this situation, synchro/simtraffic would stop every vehicle following the left turning vehicle, where in real life people would pass on right.
I would be interested to hear the fix.

Dear John and Jahid,

Thanks for your replies. I am working to see the reliability of travel time runs.

Thanks,
Ravi