An Investigation Of Centroid Connector Placement For

!
An Investigation Of Centroid Connector Placement For Advanced Traffic Assignment Models With Added
Network Detail
Ehsan Jafari; Mason D. Gemar; Natalia Ruiz Juri; Jennifer Duthie
!
!
!
!
!
!
!
!
!
!
!
!
The University of Texas at Austin
!
!
!
f
!
!
f
!
!
!
INTRODUCTION
Closest Node
CASE STUDY
Bi-level
!
!
!
!
!
!
Research Motivation
!
!
!
β€’ Advanced traffic assignment models typically incorporate more detailed
network representations.
!
!
!
!
!
!
!
β€’ Changing network topology by adding detail to the network requires carefully
editing the connector structure.
!
!
!
!
!
!
!
β€’ The number and placement of centroid connectors can also significantly
impact the congestion pattern in capacitated models.
Research Objective
!
1
𝛾𝑖 =
𝑓𝑖
𝑗
!
β€’ Two strategies are proposed for centroid connector placement.
β€’ A new metric has been introduced, the locality factor, to describe the use of
minor streets by local traffic.
APPROACHES
1- Closest Node approach:
!
𝐹𝑗
𝛿𝑖𝑗
𝑑𝑖𝑗
!
!
𝛾𝑖
𝑓𝑖
𝐹𝑗
𝑑𝑖𝑗
𝛿𝑖𝑗
!
link 𝑖 locality factor
total flow on link 𝑖
path 𝑗 flow
distance between link 𝑖 and path 𝑗
1 if link 𝑖 is on path 𝑗, 0 otherwise
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
Closest Node! Selection Layout
!
!
!
!
!
!
RESULTS
Bi-Level Selection Layout
!
!
!
!
!
!
!
!
!
Contributions
Complete Network Layout
!
An alternative to measure the share
of local traffic on each link:
β€’ The primary purpose of this study it to determine the importance of
modeling local streets and to address issues related to including them in a
dynamic framework.
Base Network
Layout
!
!
Locality Factor (LF)
β€’
!
!
!
!
!
!
!
!
!
!
!
!
!
β€’ The bi-level strategy increases the share of local traffic
on local streets because some portion of demand is
required to start/end their trip within the inner subzones
β€’ TSTT decreases as the number of connectors
increases because:
β€’ more alternative routes with lower costs are
available in a network with more connector links,
β€’ also less concentration of trip ends at any particular
connector location
β€’ LF decreases which suggests a trend toward avoiding
minor roads when alternatives are present
!
!
β€’ The effect of inner-centroid fraction on locality factor
(left) and TSTT (right) using the bi-level selection
approach
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
Total Locality Factor (TLF)
Network
Complete
Closest
Bi-Level
Total Locality Factor
0.087
0.092
0.14
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
β€’ The first approach radially distributes the connectors to link centroids to
the nearest nodes in order to account for added network detail, and is
intended to exemplify some of the limitations of the most common
techniques in practice.
2- Bi-level approach:
β€’ The second strategy involves dividing the centroid and subsequent
demand into two parts and distributing the demand across one
subcentroid linked to nearby nodes and one linked to the periphery,
effectively establishing a bi-level distribution.
β€’ Evaluation with field measurements – Williamson
County, TX,
β€’ 1,281 nodes, 158 zones, 2,617 links
β€’ Travel times along 9 corridors covering 365 links were
collected
β€’ The best performance is achieved using the bi-level
approach by assigning 90% of demand to the inner
subzones
Network
Layout
Travel Time
Error (min)
Base Network
Bi-level
Network
Closest Network
min max avg
min max avg
min max avg
.03
.03
.02
5.4
1.2
5.4
1.3
4.2
1.1
Sensitivity to the Number of Centroid Connectors
SUMMARY
!
!
!
!
!
!
!
!
!
Number of
Connectors
per SubZone
β€’ Incorporation of minor roadways without any changes
to the centroid connector layout may lead to inaccurate
usage patterns
β€’ Two strategies were proposed to redistribute the
connectors
β€’ A new performance metric was also introduced, the
locality factor, to quantify the utilization of roadway
segments by demand generated from (or attracted to)
nearby centroids
β€’ The bi-level method and selection process results in
the highest locality factor and generally achieves better
model performance metrics
!
!
!
!
!
Demand # of Centroid
Split
Connectors
TLF
TSTT (hr)
!
!
!
!
!
!
!
!
!
!
!
!
!
1
2
3
4
5
6
7
8
50-50
50-50
50-50
50-50
50-50
50-50
50-50
50-50
239
371
471
571
653
691
773
811
0.55
0.25
0.20
0.15
0.14
0.12
0.10
0.11
8748.87
8310.6
8318.21
8317.08
8153.54
8071.98
8057.52
8010.96
!
!
!
!
!
!
!
!
!
!
!
collaborate. innovate. educate.