Post on 18-Dec-2015
Junction Modelling in a Strategic Transport Model
Wee Liang LimHenry Le
Land Transport Authority, Singapore
Outline
• Background
• Objective
• Overview of the LTA Strategic Transport Model
• Review of iterative junction modelling
• Revised junction modelling
• Comparison of performance results
• Conclusions
BackgroundSingapore
• A city state
• 648 km2 area ; 4.1 mil population.
• 109 km rail lines (MRT/LRT), 150 km expressways
• 575 km major arterial roads, 1500 signalised junctions
EMME/2 Strategic Transport Model
• Used widely to forecast travel demand for planning & design of transport proposals, also calculate user benefits
• Enhanced over the years
• Incorporated “iterative” junction modelling in 2000
• Recently revised junction modelling
Objective
• To present a review of the iterative approach in junction modelling and its limitations.
• To present a revised & simpler approach in junction modelling and its improvements in model convergence
Trip Generation
Trip Distribution
Mode Split
Peak Hour Factors
Trip Assignment
Model Inputs Model OutputsModel Step
iteration
HBW (car, m/c, taxi, LRT, MRT, bus, c/o bus) HBS (car, LRT, MRT, bus, school bus) HBB, HBL, NHB
Daily OD matrices by mode and trip purpose
- travel times- highway volumes- transit volumes- other performance measures for downstream analysis(e.g. financial, economic analysis)
Model outputs
- car, m/c, taxi- LRT/MRT- c/o bus, school bus- bus
Peak hour matrices, AM, PM & OP by mode
HBW (highway, transit) HBS, HBB, HBL, NHB
Trip distribution matrices by trip purpose and main mode
HBW, HBS, HBBHBL, NHB
Daily trip ends by purpose
OVERVIEW OF LTA STRATEGIC TRANSPORT MODEL
- Planning Data: population, employment, school enrolment.- Car ownership, - Dwelling types & others
Land use data
Trip rate data
From assignments- Car, m/c, Taxi- LRT/MRT/Bus
Skims of time and cost
From HIS and traffic count data
Peak hour factors by trip purpose, mode and area
- tourist trips, airport trips- goods vehicle trips
Special trip matrices
- links, junctions- travel time, delay functions- transit services
Network
From HIS and SP survey
Mode split parameters
Trip distribution functions
HIS data
Junction Modelling - Iterative Approach
Review Standard
Calculate link delay
Assign Traffic
Start
Check Convergence
No
END
Yes
Calculate link delay
Calculate Junction delay
Calculate movement capacity & effective green time
Iterative Approach
Run assignment for N iterations
Start
Check Convergence
No
END
Yes
Assignment Procedure
Iterative Approach Review
• Turn penalty (delay) function (tpf):
• User defined turn data– UP1: 6 digits to store
1:No. of lanes 2: No. of short lanes
3:Shared lane description 4: Signal control or not
5:Opposed information 6: unused
– UP2: unopposed green time & opposed green time
– UP3: cycle time
• Extra user turn data: effective green time & capacity
Junction Coding
Iterative Approach Review
• Delay function was based on SIDRA Formulae
• Delay = uniform delay + Overflow delay
• Function of cycle time, green split, arrival flow and movement capacity
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4V/C
Del
ay (
min
ute
s)
Delay Function for Signalised Movement
D(delay) = c/2*(1-u)2/(1-u*x)
+ 900*(x-1 + Sqr((x-1)2 + 4x/C))
Iterative Approach Review
• Unopposed Movement
– Capacity = Saturation flow*green time/cycle time
• Opposed Movement:
– Opposing movement & flow
– Effective saturation flow
– Effective capacity for opposed movement
• Movement in a shared lane:
– Capacity is proportioned to the ratio of its flow over total lane flow.
Movement Capacity
Assignment & convergence instability. Factors identified:
(i) Steep junction delay curve
(ii) Iterative calculation of movement capacity
Junction Delay Function
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
V/C
De
lay
(m
inu
tes
)
Iterative Approach Review
Limitations
1st Iteration
3rd Iteration
2nd Iteration
4th Iteration
Revised Approach
• To represent realistically the junction delay in a strategic network
• To improve model convergence and therefore assignment stability and accuracy
Objectives
Junction Modelling - Revised ApproachAssignment Procedure
Calculate link delay
Calculate Junction delay
Calculate movement capacity & effective green time
Iterative Approach
Assign Traffic
Start
Check Convergence
No
END
Yes
Calculate link delay
Calculate Junction delay
Calculate movement capacity & effective green time
Revised Approach
Assign Traffic
Start
Check Convergence
No
END
Yes
Revised Approach
To reduce the steep gradient of the iterative delay curve
Junction Delay Function
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0 v/c
dela
y(m
in)
Current delay function
Revised Delay Function
Revised Delay Function
Delay = {0.25 + 0.25 (V/C)}*{c-g} for V/C <1 {0.5 + 1.5 (V/C-1)}*{c-g} 1 < V/C < 2 {2 + 2 (V/C - 2)}* {c-g} 2 < V/C
Source:V/C < 1: uniform delayV/C > 1: calibration of the base model
Revised Approach
• Different base saturation flow (veh/hour)
Left Through Right
1700 1960 1800
• Simplified calculation for shared lane movements
Saturation flow = base saturation flow/no. movements
• Added calculation for short Lane
Saturation flow = storage length/(vehicle space* mov. green
time)
(Capacity 400 veh/hr)
• Simplified calculation for opposed movement
Saturation flow = base saturation flow/3
(Capacity 200 veh/hr)
Revised & Improved Calculation of Movement Capacity
Comparison of movement delays
Left Movement
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
0.0-0.1 0.1-0.2 0.2-0.3 0.3-0.4 0.4-0.5 0.5-0.6 0.6-0.7 0.7-0.8 0.8-0.9 0.9-0.10
>1.0
Delay(min)
Perc
en
tag
e o
f Ju
ncti
on
Iterative Revised
Iterative: Ave 16.8 sec
Revised: Ave 22.2 sec
32% increase
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
0.0-0.1 0.1-0.2 0.2-0.3 0.3-0.4 0.4-0.5 0.5-0.6 0.6-0.7 0.7-0.8 0.8-0.9 0.9-0.10
>1.0
Delay(min)
Per
cen
tag
e o
f Ju
nct
ion
Iterative Revised
Iterative: Ave 30.0 sec
Revised: Ave 27.0 sec
10% reduction
Through Movement
Comparison of movement delays
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
0.0-0.1 0.1-0.2 0.2-0.3 0.3-0.4 0.4-0.5 0.5-0.6 0.6-0.7 0.7-0.8 0.8-0.9 0.9-0.10
>1.0
Delay(min)
Perc
en
tag
e o
f Ju
ncti
on
Iterative Revised
Iterative: Ave 38.4 sec
Revised: Ave 43.2 sec
12.5 % Increase
Right Movement
Comparison of movement delays
Comparison of network travel time
1999 Network - AM peak
Link Travel time Junction Delay Total Travel TimeIterative Method (hrs) 68411 13845 82255Revised Method (hrs) 67409 15717 83126Difference (hrs) -1002 1873 871% Change -1.5% 13.5% 1.1%
Observations:• Junction delay increased despite delay curve smoothened• Link travel time reduction => more efficient route choice, more converged assignment
Improvement in model convergence
Comparison of model running time on the 2015 network
The revised approach has improved model convergence through reducing number of iterations & running time.
Unix system(450 MHz)
Pentium 4(2400 MHz)
Stopping criteria
Stopping Gap 0.5 Stopping Gap 0.1
Iterative Approach 34 hrs (38) 9.5 hrs (120)
Revised Approach 23 hrs (30) 7.2 hrs (94)
Difference -32% (-21%) -24% (-22%)
Note: (38) number of iterations per highway assignment
Conclusion
• Junction delay is a major contributor to a journey time in an urban network.
• Full incorporation of SIDRA to a strategic transport model may not suitable.
• Revised and simpler approach to calculation of junction delay was presented
• The revised model represents realistic movement delays, travel times and traffic demand in a network.
• Model converges faster and predicts stable travel time & saving for transport schemes.