Presentation Materials

Analysis of the SHRP 2 Naturalistic Driving Study Data [S08(B)]
Evaluation of Offset Left-Turn Lanes
Jessica M. Hutton
Presentation to MCTRS
August 15, 2013
Research Objectives
Phase 1
• Determine if we can use the NDS dataset to evaluate
gap acceptance behavior at offset left-turn lanes
• Develop and validate a study design to do this
Phase 2
• Answer the study questions:
– Do offset left-turn lanes affect turn behavior and gap
acceptance?
– What effect does the presence of a vehicle in the opposing leftturn lane have?
• Develop design guidance for offset left-turn lanes
2
What is an Offset Left-Turn Lane?
3
Left-Turn Lanes and Sight Distance
4
Research Activities
IRB approval and
Data Sharing
Agreement
Reduce video data
Identify
intersections for
research
Define and refine
data request and
receive data
Analyze data
Draw conclusions
and develop
recommendations
5
IRB and DSA
• This research is designed specifically to avoid the
collection or use of personally identifying information of
NDS participants.
• We are exempt from IRB approval.
• We cannot use the NDS data for any other research
without another DSA, and we cannot share it with third
parties.
• If we develop any datasets that could be useful to other
researchers, those will be made available to SHRP 2.
6
Identifying Intersections
• To answer our research questions, we need
intersections with the following characteristics:
–
–
–
–
Dedicated left-turn lanes
Unsignalized or permissive left-turn signal phasing
No sight limitations due to geometry
Drivers making left turns there
• Among the intersections chosen, we need a
variety of offset conditions to observe how
changes in offset correlate with changes in gap
acceptance behavior
7
Negative offset
Unsignalized
Zero offset
Positive offset
Signalized
Highway agencies
can help identify
intersections with
desired signal
phasing
Heavily traveled intersections
Any of these
intersections could
be considered for
use in analysis
Intersections with dedicated left-turn lanes
Trips made through intersections
All NDS data
8
Data Request
• Using NDS data to identify study intersections
– CTRE has gathered roadway characteristics information for
many highly traveled routes. They are currently helping to
develop a list of intersections for us to evaluate for inclusion in
the research, and are providing the following data for each
intersection:
• Traffic control (signalized, two-way stop, etc.)
• Number/type of lanes
• AADT of approach legs (to the extent available)
• GPS coordinates
– VTTI created trip maps early in the project to identify the routes
with the highest number of trips by NDS participants. The list of
intersections provided by CTRE will be narrowed to those with
the highest trip counts.
9
Data Request
• After list of potential study intersections is
developed:
– Research team views intersections in Google Earth to look for
desirable characteristics such as:
• 90° intersection angle (no skew)
• No significant grade or horizontal curvature on approaches
• Amount of offset of left-turn lanes (need to get a range from negative to
positive)
– Research team identifies approaches of interest at each
intersection
– Research team provides VTTI with final list of intersections for
which data will be requested
10
Data Request
• Data request is location based
– We provide VTTI with the GPS coordinates of the intersections
we chose to study, and they provided us with forward and rear
video, as well as some elements of the time-series data, for any
left-turn trips made through that intersection.
– The beginning and end of a trip are defined by proximity to the
center of the intersection. A “geofence” is placed at a 500-ft
radius boundary around the center of the intersection.
– VTTI helped develop this solution after we had difficulty defining
the beginning and end of a trip in terms of time. Trips at a given
intersection all begin and end in the same location, and are of
various lengths of time depending on how long drivers wait at the
intersection and how fast they are driving.
11
Using the Data
• Our primary data source is the forward and rear video
• A Community Viewer was developed as a tool by VTTI to
view all camera views simultaneously, but this didn’t
work well for us.
• In Phase 1, data reduction was manual—we watched
forward video and rear video independently and
recorded observations in an Excel spreadsheet.
• For Phase 2, we started developing our own viewer to
watch the forward and rear video together, and are
integrating a data-reduction interface specific to our
project.
12
User Interface for Data Reduction
13
Using the video data to measure gaps
Time T0: First opposing
through vehicle reaches the
center of the intersection
after the study vehicle
arrives
Time T1: Next opposing
through vehicle reaches the
center of the intersection
Time T2: Turn is made
by study driver
T0, T1 and T2 are viewed in the forward-facing camera
Rejected Gap = T1 – T0
Time T3: First opposing
through vehicle reaches the
center of the intersection
after the study vehicle
makes the left turn
T3 is estimated from forward
camera or viewed in rear
camera
Accepted Gap = T3 – T1
14
Data Elements from Video Reduction
• For each turn made by NDS study vehicle:
–
–
–
–
–
Time when study vehicle arrives in queue
Time when study vehicle becomes first in queue
Position in queue when study vehicle arrives
Time when turn is made
Signal indication when study vehicle arrives and when turn
is made (also, time when signal indication changes if
applicable)
– Light and weather conditions
• For each gap accepted or rejected by NDS or nonNDS driver:
– Timestamp for end of each gap
– Presence of opposing left-turner and impact on sight distance
– Presence of opposing right-turner
15
Analysis Approach
Hypotheses to be Tested:
• Do different types of offset affect drivers’ gap
acceptance behavior?
• Does presence of an opposing left-turn vehicle
affect that gap acceptance behavior?
• Do different types of offset affect the rate of
occurrence for erratic maneuvers during left turns?
Statistical Method Used:
• Logistic regression analysis
16
Phase 1 Summary Statistics
Unique
object ID
(VTTI)
7
13
19
23
28
33
Number of left-turn
maneuvers recorded
Offset
type
Positive
Zero
Negative
Total
NDS
vehicles
9
1
32
25
23
3
93
Non-NDS
vehicles
2
0
10
1
14
0
27
Total
Total
number
number of
of
rejected accepted
gaps
gaps
39
2
4
0
87
10
34
2
104
15
10
1
278
30
Number of events
with a vehicle
present in the
opposing left-turn
lane
No
20
4
61
31
81
11
208
Yes
21
0
36
4
20
0
81
17
Distribution of Gaps by Left-Turning Study Vehicles
18
Logistic Regression Models by Offset Type
19
Anticipated Phase 2 Research Results
• Design guidance and recommended criteria for
offset left-turn lanes
• Recommended updated text for left-turn lane
discussion in Chapter 9 of the Green Book
• Useful to traffic safety engineers, design
engineers, and planners in state and local
highway agencies, as well as consultants and
researchers
20