Transcript Slide 1

Estimating the Traffic Flow
Impact of Pedestrians With
Limited Data
Daniel Tischler, Elizabeth Sall,
Lisa Zorn & Jennifer Ziebarth
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
14th TRB Planning Applications Conference
May 6th, 2013
Making the Case
I have problems.
I’m trying to model San Francisco traffic conditions, but
Pedestrians affect vehicle capacity
My dynamic traffic assignment model is needy
Data is scarce
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
2
Network capacity
Road capacity is a key input in traffic assignment
We assign capacities to roads by facility classification schemes
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
3
Network capacity
In our DTA model, signal timing becomes the primary determinant of capacity
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
4
Pedestrians interact with vehicles
Lots of pedestrians crossing the street prevent cars from turning
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
5
Pedestrian volumes vary
Our meso-level, dynamic assignment model does not simulate pedestrians.
It does not know that right turns at “A” are more restricted than at “B”
Intersection A
Intersection B
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
6
Methodologies
Analytical methods
Simulation methods
Local observations
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
7
Analytical approaches
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
8
Analytical approaches
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
9
Pedestrian volume – vehicle flow relationship
Right Turn Saturation Flow Rate (HCM 2010)
Saturation Flow Rate
(veh / hr)
2000
1500
1000
500
0
0
500
1000
1500
2000
2500
Pedestrians crossing Street (peds / hr)
•
Assumes 2 sec veh-veh headway, 50% green time
•
Doesn’t allow for ped volumes > 5,000
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
10
Comparing different methods
HCM2010
Rouphail and Eads, 1997
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
11
Applying pedestrian friction in the model
Saturation flow rate
Green time
Follow-up time
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
12
Scale of Application
Area Type Method
Neighborhood Type
Method Method
Uniform Parameters
Unique Intersection
Method
55%
Existing approach
74%
Under
consideration
Do nothing
approach
87%
Under consideration
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
13
Resources
SFMTA pedestrian count program
SFMTA pedestrian model
Project-related pedestrian counts
Pedestrian-vehicle observations
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
14
SFMTA pedestrian count program
•
2-hour pedestrian counts at 50
locations (2009/2010)
•
Six automatic pedestrian
counters (24/7 observations)
Actual pedestrian count binder
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
15
Automated pedestrian counters
•
Time and day profiles
Tenderloin
Union Square
Castro
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
16
Pedestrian count locations
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
17
HCM 2010 flow rate adjustment factors
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
18
SFMTA pedestrian model
Used count data and log-linear
regression model to estimate
pedestrian volumes throughout
San Francisco
lnYi = β0+ β1 X1i+ β2 X2i+ … + βj Xji
•
•
Combined with auto traffic and
collision data to develop
pedestrian crossing accident
risk factors
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
19
Pedestrian volume estimates
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
20
Local validation of pedestrian estimation
Ped Counts
12,000
10,000
8,000
6,000
4,000
2,000
Market
Mission
0
8th
7th
6th
Howard
5th
4th
Folsom
3rd
PM Peak Hour Ped Crossings
PM Peak Hour Ped Crossings
Ped Model Estimates
12,000
10,000
8,000
6,000
4,000
2,000
Market
Mission
0
8th
7th
6th
Howard
5th
4th
Folsom
3rd
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
21
Pedestrian-vehicle interaction observations
SFCTA staff observed downtown
intersections of varying pedestrian volumes
Collected information on quantity of
pedestrians and relative restriction of
vehicle turning movement capacity
The urban environment is complex:
•
Groupings of peds, directions of travel
bicycles, unique timing plans, variations
in geometry, etc.
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
22
Observed pedestrian-vehicle interactions
Percent reduction in Vehicle Capacity
Pedestrian Impedance of Turning Vehicles
120%
Incremental pedestrians have
the greatest impact at low
pedestrian volumes
100%
80%
60%
40%
20%
0%
0
500
1,000
1,500
2,000
Pedestrian Flow in Crosswalk (peds/hr during walk phase)
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
23
Observed pedestrian-vehicle interactions
HCM2010
Rouphail and Eads, 1997
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
24
Model Tests
Uniform parameters method
Area type method
Neighborhood type method – sat flow rate
Neighborhood type method – green time reduction
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
25
Focus on pedestrian count-rich area
SoMa pedestrian
count locations
1. 4th / Market
2. 6th / Market
3. 8th / Market
4. 6th / Mission
5. 3rd / Howard
6. 7th / Folsom
1
5
2
4
3
6
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
26
Hourly pedestrian counts
3,000
3,300
3rd St
5th St
7th St
Market St
10,000
Mission St
1,100
Howard St
1,200
Folsom St
600
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
27
Network parameters
Uniform parameter method
All
Signals
Sat flow
1,800
Spacing
2.0s
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
28
Network parameters
Area type method
Hi Peds
Sat flow
1,620
Spacing
2.22s
Med
Peds
Sat flow
1,710
Spacing
2.11s
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
29
Network parameters
Neighborhood type method
Adjust saturation flow rates
Sat Flow
810
Spacing
4.4s
Sat Flow
1,260
Spacing
2.9s
Sat Flow
1,530
Spacing
2.4s
Sat Flow
180
Spacing
20s
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
30
Network parameters
Neighborhood type method
Reduce green time for turning movements
55% less
green
90% less
green
30% less
green
15% less
green
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
31
Testing pedestrian friction methods
•
•
Evaluation not yet complete!
Numbers
Area type method
At select subarea nodes,
turning movement volume
validation improves with
neighborhood type method
•
Queuing
Neighborhood type method
Neighborhood method seems
to produce more realistic
queue formation than area
method
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
32
Conclusions
•
•
•
•
Pedestrian-vehicle friction is very important at some
locations!
Numerous options to make assignment models
sensitive to pedestrians
The most realistic approaches are very difficult
We still have work to do:
•
•
•
•
•
Run more tests
More local validation
Develop better neighborhood system
Incorporate time of day factors
Develop strategy for scenarios that change
pedestrian volumes
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY
33
The End.
Questions?
SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY