I need to calibrate a network model in Simtraffic based on the travel time I got from the field. Right now, the travel time I got from Simtraffic arterial report is much smaller than the field data for most of the links. However, some links give me a very close result. So I guess I need to calibrate the model on a link basis. And I feel that it's the commercial accesses adjacent to the roadway cause the big difference between output travel time and field data. I didn't put all the minor driveways in my model, but they do generate a lot of left-in/out traffic that will greatly affect the operation for a two-lane two way corridor. Does anybody have some similar experiance in calibration for a situation like this before? Any suggestion is appreciate.

Jingjing

This is probably a stupid question on my part, but have you set the link speeds to what you observed in the field?

As for the driveways... you may just want to include ones with high volumes. Like at superstores, malls, etc.

Mark:
I'm trying not to put a very big change on the link speed just for thoes problem links. So I'm wondering is there any other parameters I can change besides the speed. In the model, I've included all the major side streets. But there are a lot of commerical accesses between the nodes that I cannot show in the model. I don't now if there's a good way to reflect that situation.
Jingjing

You could also try adjusting the turning speeds, but the defaults are "usually" pretty realistic.

Have you looked at how your signals and offsets in Synchro compare to the field? Your signal progression may be different.

You can also look at the Driver and Vehicle Characteristics under the Options menu in SimTraffic. I wouldn't go crazy changing the defaults in there, but you might see something that doesn't fit your job. Tweaking these would be my last resort.

Thanks for the suggestions. I didn't check the timing in the field. All the phasing, timing and coordination information are based on the timing plans got from the county. I will go ahead check the driver and vehicle parameters you mentioned. Are they network wide adjustments?
Jingjing

yes, any changes to the driver or vehicle parameters table will be network-wide.
The driveways may be an issue - it there's one with turning volume significant to the network, then you may want to consider including it in the network. Those cars will accelerate and decelerate to turning speeds, where vehicles that dissappear or appear in the network (network imbalances) do so "at speed"

Eric

Eric:

Thank you for the suggestion.

Jingjing

Hi all,

Mark, Eric and Jahid have posted some great comments (as they always do). Thanks. I'll add a bit to the conversation.

Before calibration, there is the verification process. That is, the model is only as good as the data (yeah, i'm stating the obvious). But, as discussed in the posts, signal timing will need to be confirmed. Paper copies of signal timing are often out-dated and/or someone changes some timing in the field but doesn't update the files at the office. A simply change in one offset at an intersection can have a dramatic impact on your travel times.

Regarding calibration, I often think of this in three main categories. This would be saturaiton flows, volumes and signal timings. There are others (especially when you are talking simulation), but this is where you would start.

On saturation flow, there are numerous factors that can influence this (such as grade, lane width, etc.). In SimTraffic some of these factors will influence the Headway factor. For instance, if you change the lane width you'll see the headway factor change for calibration. So, if flows are a concern, you'll want to do some field saturation flow studies and compare to the models. As needed, adjust the factors to achieve the field represented result.

The volumes can be a very big source of problem and I'm not just saying the errors in counts. It has to do with service vs. demand. When most count, they count as the vehicle passes the stop bar. This would be service volume. In reality, the demand could be higher so you might want to capture the leftover vehicles after each cycle (or at the end of a period). In your case, you say the simulation is better than the field, so this may not be the issue.

For signal timings, this is as much verifying the field timings as it is calibrating. Make sure you've acurately entered all the parameters (controller type, yeild points, recalls, etc, etc, etc). These factors can have an impact on how vehicles are leaving a coordinated (actuated-coordinated) signal.

You also mention driveways could be the source, and I tend to agree based on what you said. You also note that you shouldn't model every little driveway, and that is also true (see the FHWA traffic signal toolbox series). But, you might include some major driveways and including turning movements in and out. This will impact the flow of traffic along your route in SimTraffic.

I guess the bottom line is there are many factors that can influence your model/result. As you are doing, it has to be narrowed down to the suspect factors. But, only the analyst familiar with the network is capable of doing this.

Good luck and thanks,

John.

Thanks a lot for all the information. That helps me a lot. I'll try to get more field information to help me calibrae the model.

This is just a follow-up question to JJF. I am having similar issues with Calibration of Simtraffic model. I am wondering what are the MOE's that I can use to compare with field travel time runs. Thanks,

Tejas Mehta