12th TRB Conference on Transportation Planning Applications

1
Automatically Balancing Intersection Volumes in A
Highway Network
12th TRB Conference on Transportation Planning
Applications May 17-21, 2009 Presenters Jin Ren
and Aziz Rahman
12th TRB Conference on Transportation Planning
Applications May 17-21, 2009 Presenters Jin Ren
and Aziz Rahman
12th TRB Conference on Transportation Planning
Applications May 17-21, 2009 Presenters Jin Ren
and Aziz Rahman
12th TRB Conference on Transportation Planning
Applications May 17-21, 2009 Presenters Jin Ren
and Aziz Rahman
2
Presentation Outline

• Need for Balanced Volumes
• Current Balancing Techniques
• New Automatic Balancing Techniques
• Formation of Intersection Turn Matrix
• Doubly Constrained Method
• Successive Averaging or Maximizing and Iterative
  Balancing
• Statistical Comparisons of Methods
• Conclusion

3
Need for Balanced Volumes

• Existing base highway network simulation in
  Synchro and VISSIM
• Unbalanced upstream and downstream post-processed
  future flow
• Build simulation confidence in audience
• Ensure simulation model run results not wacky
• Take into account mid-block driveway traffic in
  simulation

4
Current Balancing Techniques

1. Manual Adjustment match the volumes departing
   one intersection to those arriving at the
   downstream intersection, or vice versa
2. EMME Demand Adjustments create a trip table and
   run traffic assignment based on intersection
   volumes
3. VISUM T-Flow Fuzzy Technique create a trip table
   to emulate intersection turning volumes

5
Pros and Cons of Each Technique

• Manual Adjustment
• a) uses a simple spreadsheet or Synchro b)
  time-consuming if numerous balancing iterations
  required
• 2. VISUM T-Flow Fuzzy Technique emulate turns
  with balanced volumes, but intra-zonal traffic
  causes turning volume losses

6
T-Flow Fuzzy Example 1
7
T-Flow Fuzzy Example 2
8
Why Introduce New Methods?

• Develop a statistically sound technique
• Reduce labor time on balancing
• Generate more accurate turning volumes
• Create an automatic process which is
  user-friendly and affordable
• Build confidence in simulation with the balanced
  volumes

9
New Automatic Balancing Techniques

• Successive Averaging/Iterative Balancing
  iteratively average downstream and upstream link
  volumes and then balance intersections
• Successive Maximizing/Iterative Balancing
  iteratively maximize downstream and upstream link
  volumes and then balance intersections

10
Formation of Intersection Turn Matrix
11
Doubly Constrained Balancing Method
ai and bj adjustments made to each O-D pair
volume in order to achieve the target values Oi
and Dj required by the growth factors for the
origins and destinations
-Factors for origins (in) and destinations
(out) -Bi-Proportional Algorithm
Formula
tij
ai
Algorithm assumption
bj
12
Schematics to Intersection Balancing
ai and bj adjustments made to each O-D pair
volume in order to achieve the target values Oi
and Dj required by the growth factors for the
origins and destinations
Err lt 0.001
No
Yes
13
Equations for Intersection Balancing
ai and bj adjustments made to each O-D pair
volume in order to achieve the target values Oi
and Dj required by the growth factors for the
origins and destinations
Doubly constrained
mth Iteration Row wise
mth Iteration Column wise
14
Successive Averaging or Maximizing and Iterative
Balancing Diagram
Non Balanced Vol.
Avg. Link level In Out Vol.
Form Intersection Turns Matrix
New Turn Vol.
Balance Intersection In Out Vol.
Apply Doubly Constrained for Turns Vol.
Adjustment
Calculate Error
Yes
Errorlt0.001?
Balanced Vol
No
Error Change?
Yes
No
15
Layout Unbalanced Intersection Volumes
Assumption Averaging in/out link volumes are
supposed to be equal.
16
Doubly Constrained Balancing
Method doubly constrained intersection arrivals
and departures
17
Example 1 Balancing Statistics
T-Flow Fuzzy Technique
Successive Average Technique
18
Example 2 Balancing Statistics
T-Flow Fuzzy Technique
Successive Average Technique
19
Statistical Comparisons
TESTS R2 RMSE Slope Mean Rel Err VOLUME DELTA
T-Flow Fuzzy Ex 1 0.96 20 0.95 12 -1358 (-3.0)
SA/IB Ex 1 0.97 17 0.96 10 4
T-Flow Fuzzy Ex 2 0.97 21 1.00 12 -1114 (-2.5)
SA/IB Ex 2 0.99 12 0.98 7 0
Findings SA/IB Example 1 and Example 2 are both
better than T-Flow.
20
Conclusion

• An innovative mathematical method is presented
  with two practical examples
• Successive averaging/iterative balancing
  technique shows better goodness of fit statistics
• Automatic balancing technique saves time in
  traffic simulation process
• The spreadsheet method can be implemented
  cost-effectively
• Capacity constraint can be incorporated in the
  balancing algorithm in future