You are on page 1of 21

Development of SMRT Strategic Public Transport

Model of Singapore
Jeffrey Ang
Email: Jeffreyang@smrt.com.sg
• Transport Simulations in SMRT

• SMRT Cube Strategic Model

• Model Development and Structure

Overview • Landuse and Distance Fares Model

• Model Performance

• Integration of Cube with Genetic Algorithm


Transport Simulation in SMRT

First SMRT Strategic Update of Strategic SMRT decided to built in-


Model using Emme2 Model and conversion house capability in model
software. of Emme2 to Cube development.

Commence update of 2007


2002

2010
2007
North East Circle Line
model developed by
Line Study Study
external consultant

EMME/2
Transport Simulation in SMRT

Decide to abandon the First in-house Second model


existing model model completed upgrade

New Features: Commence second overhaul


2011

2012

2014
Team decided to build
CUBE : New 4-Stage Model of the model to improve
model from scratch
CUBE : New AM Peak model to model PPHPD forecasting horizon:
Python : New Python model to calculate Distance Fares URA MasterPlan 2014
ArcGIS : New land use model for Trip Generation Cube Version Upgrade to 6.0
Forecast horizon extended from 2016 to 2021
Transport Simulation in SMRT

Second model Purchase of: Purchase of: To commence 3rd


upgrade MassMotion (Station Modelling) OpenTrack (Trains Simulation) Cube model update
completed Vissim (Traffic and Mobility Simulation)

To update Cube land use

2020
2016
Expanded capability into:
2015

New Features: Takes on external consultancy projects

2018
Migration of CUBE 5.0 to 6.1 Trains simulation using OpenTrack model based on the latest
Expanded capability into: simulation software URA Master Plan 2019
URA MasterPlan 2014
Zones increased from 1180 to 1503 Pedestrian and Station Simulation
Nodes increased 4.5x (5400 -> 24,600) Traffic and Mobility Assessment
Links increased 3.0x (20,300 -> 61,800)
Forecast horizon stretched from 2021 to 2030
Overview – Transport Planning & Modelling in SMRT
• Cube Voyager (Public Transport Ridership Modelling)
• Long term forecasting for public transport network

• VISUM (Master Plan)


• Simulation of private and public transportation

• VISSIM / SIDRA (Traffic Simulation)


• Traffic simulation, Traffic Impact Assessment and Roundabout capacity and performance analysis.

• MassMotion (Pedestrian & Station Simulation)


• Crowd and flow analysis for buildings, events and evacuation.

• OpenTrack /OpenPowerNet (Railway Simulation)


• Capacity of lines and stations, robustness of timetable, changes to infrastructure

• Others: ArcGIS / Python (Connectivity and Accessibility)


• Walkability Analysis, Network Connectivity of study area
Transport Services Provided by SMRT

Transport Modelling
Transport Master Plan
and Simulations

Station and Pedestrian Traffic Impact


Simulations Assessment

Mobility and Public Transport


Accessibility Study Planning

Railway Simulation Highway Engineering


Cube Voyager
SMRT Strategic Public Transport Model for Singapore
Thomson East Coast Line

North East Line


North South Line

Jurong Region Line

East West Line


Downtown Line

Circle Line
Model Development and Structure
A. Calibration of Base Year Model B. Forecast for Future Years
Landuse
Model
Landuse Model
1
Sources of
Landuse Data

2 SMRT GIS Landuse Model

• Split into 15,000 landuse polygons

• Clustered into 7 Regression Groups

• 10 landuse layers from year 2020 to 2030


3 Cube 1503 Zones
RIDERSHIP BY LINE : TRIP LENGTH :
+/- 10% 5% - 10%

Model
Performance
Cube Forecast (~3 years before line opens)
Vs Actual (6-12 months after line opens)
PASSENGER PER HOUR PER LOCATION OF PPHPD:
DIRECTION (PPHPD) : 100%
+/- 5%
Modelling of Distance Fares
A. Journey Distance Based Fares B. Estimation of Distance Fares in Cube

• Journey based calculations

• Split by fare type

• Apportionment of revenue to services


Future
Development
Integration of Cube with Genetic Algorithm
Network Planning using Genetic Algorithm

Genetic Algorithm reflects the process of


natural selection where the fittest individuals
are selected for the reproduction in order to
produce offspring of the next generation.
Search for all possible routes subjected to pre-defined rules &
conditions, to identify the best solution which minimizes resources

Sample: GA Route Representation


R1 R2 R3

Route Representation Route Set

in Genetic Algorithm

• Network consist of route segment 1, 3, 5, 7 and 9


Red Line

Sample Results
Scenario 1: RED & BLUE LINE

*Showing the route for Red Line only for illustration

Red Line Blue Line


Estimated Demand : 391 Estimated Demand : 305
Bus Capacity : 24 Bus Capacity : 24
Required Trips : 16 Required Trips : 13
Headway : 4.0 minutes Headway : 5 minutes
No. of Buses : 10 No. of Buses : 6 No. of Buses : 16
Integration of G.A. and Cube Network Preparation in Cube

Import to NetworkX

Coding of Road Network in NetworkX Network in NetworkX

Generation of Feasible Routes Sets Generation of Feasible Routes Sets


Genetic Algorithm in Python

Genetic Algorithm in Python


Initial Population Initial Population

Fitness Evaluation Fitness Evaluation

Select the best solution to the next Select the best solution to the next
generation generation
Develop a process to convert
NetworkX to Cube Network
Select good parents for breeding (vice versa) Select good parents for breeding

Fitness Evaluation Fitness Evaluation


(Light Weight Assignment Module) (Light Weight Assignment Module)

Manual re-code Export to Cube

Cube Model with New Network Cube Model with New Network

A. Existing Process in Python B. Common Network Representation in NetworkX and Cube C. Integrated Process GA and Cube
Thank you

You might also like