Problem 5: Network Simulation
Printable Version
This problem
demonstrates how a network simulation model can be used to augment studies
conducted with HCM methodologies. Simulation models offer the advantage of
being able to examine networks of highway facilities in a highly unified,
holistic fashion. Inter-dependencies and cascading effects can be taken into
account as can traffic variations of time, over saturation, queue length
fluctuations, lane blockages, and other transient phenomena. The only major
drawback is that simulation models are typically data-hungry and they take
time to develop, debug, calibrate, validate, and run. Distilling the results
also takes time, because there’s so much information to study, absorb, and
comprehend.
Two main decisions need
to be made: 1) what network to analyze and 2) what traffic volumes to use. Exhibit 4-77 shows the
network used as the basis for the simulation model. It encompasses Alternate
Route 7 and the interchange complexes at either end: Exits 6 and 7 on I-87,
the interchange with Route 9 on Route 7, and the interchanges with Route 7
and 23rd Street on I-787. What it doesn’t include is the underlying surface
arterial network and the freeways that lie outside the artificially defined
boundary. The actual simulation network is detailed, with information about
lane configurations, vertical and horizontal geometry, speed limits, etc.
The time period we
studied was the AM peak. Either the AM peak or the PM peak would be a good
choice. The only difference is the direction of peak flow. In the AM Peak,
the flows are predominantly southbound and eastbound.
| What are the characteristics of the network that you
would expect to input into a simulation model? |
| What are some assumptions we might make for this
specific network? |
Discussion:
Take
a few minutes to consider these questions. When you are ready to continue,
click continue below to proceed.
[
Back ] to Problem 4 [
Continue ] to Sub-problem 5a |
Page Break
|
Exhibit
4-77. Simulation Study Network
|
Page Break
Sub-problem 5a:
Network Simulation
Step 1. Setup
VISSIM is the traffic
simulation package we chose to use, so the specific input data files reflect the
needs of that software. However, many of these characteristics and inputs
would be required for other modeling software as well. A summary of the inputs
and assumptions we used is as follows:
|
network
configuration (highway sections and their connections) |
|
link distances set
to values obtained from the local MPO (Metropolitan Planning Organization)
accurate to the nearest 10 feet |
|
geometrics for all
highway segments |
|
freeway free-flow
speed density functions set to a minimum of 50 mph and a maximum of 60 mph |
|
ramp speed density
functions for the loop ramps, etc., set to a minimum of 22 mph and a
maximum of 28 mph |
|
ramp speed density
functions for the right-hand ramps and semi-direct ramps set to a minimum
of 43 mph and a maximum of 48 mph |
|
ramp speed density
functions for all remaining facilities set to a minimum of 27 mph and a
maximum of 35 mph |
|
trip matrix using
locations A through M in Exhibit 4-78 as
the origins and destinations |
|
heavy vehicle
percentages set nominally at 5% throughout the network |
The O-D trip matrix
was obtained from the Capital District Transportation Committee (CDTC), which
serves as the local MPO. The matrix is derived from the outputs of the
travel forecasting model that CDTC uses in all of its planning studies.
Since it is generated data rather than observed volumes, we do not expect
exact matches between the traffic volumes observed in the field and those
predicted by the O-D trip matrix. The traffic volume on a given link in a
given time period predicted by the O-D trip matrix may be different from the
value that was actually observed in the field. The difference is that the
predicted value uses a routing algorithm to assign flows to network paths
that might not be consistent with the way drivers choose routes.
Consider:
|
What type of output or measures of effectiveness might
we expect to obtain from simulation models that would provide value above
and beyond using HCM methodologies? |
|
How does a
simulation model help develop engineering solutions? |
Discussion:
Take
a few minutes to consider these questions. Click continue when you are ready
to proceed.
[
Back ] [
Continue ] with Sub-problem 5a |
Page Break
|
Exhibit
4-78. Origins and Destinations for the Trip Matrix
|
Page Break
Sub-problem 5a: Network Simulation
Step 2. Results
The most interesting
result from the simulation model is predictions of travel times through the
network. Exhibit 4-79 presents a selected set of those values. You can see
that it takes 467 seconds (seven minutes and 47 seconds) to go from a point
300 feet south of the 23rd Street off- ramp to a point 2,600 feet
north of the merge between I-87 and the right-hand ramp from NY-7 to I-87.
These values are valuable in indicating where travel times are within
acceptable ranges and where they are not. Geometric improvements may help
some O-D pairs and not others or alleviate congestion in one place and
create it in another. These travel times help to keep track of those impacts
and relationships.
Another output that’s
particularly valuable is speeds through the weaving sections. Weaving
sections tend to be bottleneck locations. Exhibit 4-80 shows
speeds for weaving and non-weaving vehicles at a number of locations in the subarea network, based on the simulation model outputs. Data for locations
1-7 are shown in
Exhibit
4-80. The slowest speeds are on Alternate Route 7 at
the I-787 interchange on the collector-distributor road between the loop
ramp from I-787 south and the loop ramp to I-787 north. The rest of the
speeds are in the 50-60 mph range, indicating acceptable operation.
[
Back ] [
Continue ] with Sub-problem 5a |
Page Break
|
Exhibit
4-79. Predicted Travel Times
|
Page Break
Sub-problem 5a: Network Simulation
Discussion
Probably the most significant lesson learned from Problem 5 is that it is
possible to develop a simulation model for the subarea network. We also
learned that it is possible to obtain interesting pieces of information
about the performance of the network to help planners and designers
understand where there are problems and what can be done about them.
The simulation model
is a diagnostic tool, not a solution generator. It can tell how a given
network configuration performs and let you compare and contrast one solution
with another. The model won’t tell where to add capacity or how much. These
need to be obtained through engineering judgment or trial and error. But you
have to develop the solution ideas.
Simulation models have
value. They can examine networks of highway facilities in a highly unified,
holistic fashion. Inter-dependencies and cascading effects can be taken into
account, as can traffic variations over time, over saturation, queue length
fluctuations, lane blockages, and other transient phenomena. Simulation
models add value when these issues are important and the interrelationships
among the facilities have to be captured. |
[
Back ] [ Continue ] to HCMAG Home
|
Page Break
|
Exhibit 4-80. Speeds
in Weaving Sections from the Simulation
|
|