Travel Time Estimation on Arterial Streets

1

• Travel Time Estimation on Arterial Streets
• By
• Heng Wang, Transportation Analyst
  Houston-Galveston Area Council
• Dr. Antoine G Hobeika, Professor Virginia Tech

2
Outline

• Objective and background
• Focusing methodology development
• Methodology validation
• Conclusion and future study
• Q A

3
Objective

• Methodologies were prepared for the proposal for
  real-time travel time estimation on major
  arterial streets.
• Requirements
• Short time interval update for real-time
  estimation
• Simple-computation time
• Make good use of real time detected traffic
  information
• Well behaved

4
About the Methodology

• The developed methodology is presented into two
  sections
• 1. Travel time estimation on an isolated
  arterial link
• 2. Travel time estimation on a signalized
  arterial link that also considers the traffic
  situation on the upstream and downstream
  links(Network Algorithms).

5
Section 1- Travel time estimation on an isolated
arterial link --Travel Time Components

• Travel time(HCM)link travel time intersection
  control delay
• Components of intersection control delay
• 1) Uniform delay
• 2) Incremental delay (over-saturation delay)
• 3) Initial delay

6
Intersection Control Delay (HCM2000) and its
weakness in short time period update situation

• Uniform Delay
• Incremental Delay
• Initial Delay

7
Developed Algorithms--Intersection Control Delay
-Observed Vehicle Group Identification
8
Developed Intersection Control Delay Algorithms

• Case 1-where there is no initial queue for the
  observed vehicle group
• Case 2-there is an initial queue for the observed
  vehicle group and its clearance time is less than
  a cycle length
• Case 3- where initial queue clearance time (d3)
  is greater than a cycle length.

9
Intersection Control Delay- Case 1 no initial
queue
10
Intersection Control Delay-Case 2 an initial
queue exists and it is smaller than one cycle
length( 0ltd3ltCL)
g1d3-r
situation
11
Intersection Control Delay-Case 3 -Initial Queue
clearance time d3 is greater than one cycle
length (d3gtCL)
12
Validation of Intersection Control Delay
Algorithms

• An intersection at N Franklin St/Peppers
  Ferry RD in Christiansburg, Virginia was selected
  to initially conduct control delay analyses based
  on traffic volume and the arrival of vehicles in
  the observed group.

13
Validation of Intersection Control Delay
Algorithms
MAE for developed algorithm result with real
control delay 10.85sec MAE for HCM2000 algorithm
result with real control delay14.28sec
14
Validation of Intersection Control Delay
Algorithms
ANOVA Table for Actual Delay vs HCM2000 results
ANOVA Table for Actual Delay vs Developed
Algorithm results
15
Total Travel Time Computation

• Travel Time Without initial Queue
• Travel time with an initial queue but without
  blackout
• Travel time with blackout (i.e. QLgt LTD)

16
Section 2- Network Algorithms

• Network conditions that influence input
  parameters
• Bottleneck on the downstream link
• Change intersection capacity
• Blackout Situation
• Change the identification of the
  observed vehicle group.

17
(No Transcript)
18
Algorithm 1(No blackout)
Is departing rate from link i smaller than
downstream links capacity?
Yes
No
Use intersection capacity of link i
Use downstream lane capacity as the intersection
capacity of link i
19
Algorithm 2(Determining the intersection capacity
of link i when blackout is on the downstream link
i1)
Is Li1 -QLi1lt100ft? (High congestion
downstream?)
Yes
No
Use the detected flow rate from downstream
detector as the intersection capacity of link i
Algorithm 1
20
Algorithm 3(Determining incoming volume when
blackout is on link i)
Is Li Qligt100ft High congestion on link i?)
Yes
No
Use the dissipated volume from link i-1 as the
incoming volume to link i
Use the smaller of the following two values a)
the dissipating rate from link i-1 b) the
intersection capacity of link i which is the
maximum dissipating rate of link i
21
Algorithm 4 (Where no detectors are available
beyond this link)
22
A Simulation from CORSIM
23
Results
24
Conclusion and future study

• Algorithms in section 1 provide accurate results
  when compared with HCM2000 by using real world
  data
• Algorithms in section 2 are robust when compared
  with CORSIM simulation results
• Real world data would be collected to validate
  the section 2 of the developed methodology.

25
Questions?