I am running simulations on a fairly congested network with frequent, but actuated-coordinated signals, and the task is to see how much more traffic the network can handle with a new development nearby. The define point of completely saturated can be quite objective, so I figure taking the Travel Time value on the main arterial would be a good guidance. I defined the network is completely saturated when the Travel Time along the arterial are increased by certain percentage or when it remains constant.

After running multiple scenarios, with adding volumes being the only changes in the whole model and keeping everything else constant, the Travel Time results were a surprise. I have a background travel time of 9.3 minutes. After added 50 vehicles, I got a surge to become 12.3 minutes; and the travel time fell back down to 10.2 minutes after added 60. As the volumes increases, up to 250 vehicles, the travel time just keep fluctuates up and down within two minutes between 10 to 12 minutes. Since there isn't much change after adding 250 vehicles, I think I can say the arterial is saturated or near-saturated even in the background scenario.
But the problem I have is I can't explain why the travel time fluctuates up and down with only volume changes, and a higher volume scenario may result a shorter travel time. Does anyone have an good explanation in terms of this travel time fluctuating observation?

Thanks for your input in advance.

suggesstions:
calibration of the existing network. timing, drivers paarmeters, signal timing, roadway network, and travel time, multiple runs of for intersection, network, approaches are increasing the way you are increasing your base volume?

If you have already done all those, my first point would be after a certain increase in volume, the progression may work better thus improve the travel time.

but travel time up and down with volume increase is kind of odd.

Hi Eric,

Jahid has hit some great points. To echo the comment, you want to make sure first that this is not simply due to randomness of the model. Many runs of each scenario with different seed numbers will be required.

Now, I think that even with that you could still see this happen. One thing to keep in mind is that the travel time is only measured for vehicles that can get into the network. What if by adding volumes you are creating more congestion at some of your intersections. That is, say that the side street now holds the green longer and more vehicles are denied entry? If they are denied entry, they can't get into the network to have their travel time captured.

Note: Denied Entry vehicle delays DO get added with the downstream intersection.

I think a valuable set of measures for your study would be denied entry and % volume simulated. When you add more vehicles, how much did the denied entry go up? You may want to look at this for the overall network and not just individual intersections.

Thanks,

John.