31st March 2009 One-day Conference on Traffic Model ling 1 Modelling very large Transport Systems Joan Serras Department of Design, Development, Environment and Materials The Open University
Dec 21, 2015
31st March 2009 One-day Conference on Traffic Modelling 1
Modelling very large Transport
Systems
Joan SerrasDepartment of Design, Development,
Environment and Materials
The Open University
31st March 2009 One-day Conference on Traffic Modelling 2
Presentation outline Introduction: a Multilevel Representation
on transport systems The TRANSIMS modelling system and its
modules A simulation of Milton Keynes using
TRANSIMS Conclusions and further work
31st March 2009 One-day Conference on Traffic Modelling 3
Introduction The role of subsystems is essential on the
behaviour of very large areas Transport network models available which can
address such areas (~106 inhabitants) These models represent the road network at one
level TRANSIMS is not an exception A methodology has been implemented to
generate a multilevel representation using a simulation of Milton Keynes with TRANSIMS
31st March 2009 One-day Conference on Traffic Modelling 4
The TRANSIMS modelling system Developed in Los Alamos during 1990s Forecast the travel behaviour of a study
area: information on traffic impact, congestion and pollution
Relevant studies: First study (1997): metropolitan region within
Dallas (~200,000 travellers) Portland Study (2002): ~1.5 million travellers Swiss study (2004): morning peak simulation
(~1 million trips) – 7.2 million inhabitants
31st March 2009 One-day Conference on Traffic Modelling 5
The TRANSIMS modelling system Microscopic approach: travel demand
estimated at the person level “synthetic population”: a virtual representation
of all the individuals living in the study area Activity-based demand rather than trip-based
Urban activity locations defined at the household level
Output of the person movement on a second-by-second basis (24h simulation)
Parallel computing
31st March 2009 One-day Conference on Traffic Modelling 7
A simulation of Milton Keynes using TRANSIMS
Purpose of the study: Can we get the data to build a multilevel representation
from the TRANSIMS output? Check its functionality in our system (cluster at the OU) Can we adapt it to simulate a non-US city? (synthetic
population generation constraints) Significant output?
Constraints: Prime use of the software in UK lack of time (PhD period) Lack of resources: only me!
Due to constraints: many assumptions were done
31st March 2009 One-day Conference on Traffic Modelling 8
A simulation of Milton Keynes using TRANSIMS Facts about Milton Keynes population (Census
2001): Population: ~200,000 inhabitants (urban area: ~170,000
inhabitants) Commuters (~60,000 commuters):
22,000 people commuting outside Milton Keynes (mainly to London area)
39,000 people commute to Milton Keynes
The Milton Keynes road network: A road grid (10 “horizontal” x 11 “vertical roads”) 1km2 each grid for easy access between them ~300 roundabouts GIS representation: 2630 nodes and 3457 links
31st March 2009 One-day Conference on Traffic Modelling 10
A simulation of Milton Keynes using TRANSIMS
Milton Keynes network From NTFS format to TRANSIMS format No traffic lights, no public transport
The synthetic population (Census 2001) US Census incompatibility: new method implemented
Household structure (150,000 inhabitants) Commuters (26,000 to MK; 13,000 out of MK)
Activity Generation survey from Balcksburg, VA (lack of time – not that different:
work, shop, visit activity types kept)
Feedback 50 iterations between Router and Microsimulator
31st March 2009 One-day Conference on Traffic Modelling 11
A simulation of Milton Keynes using TRANSIMS
Clips on the Milton Keynes model can be seen in the following website: http://design.open.ac.uk/serras/miltonKeynes_simClips.htm
31st March 2009 One-day Conference on Traffic Modelling 12
Conclusions and further work A simulation of Milton Keynes using
TRANSIMS has been produced at the OU Fairly good results have been produced Significant margin for improvement
Currently working on improving the model Data has already been used on a two-level
representation More levels need to be defined in order to infer
relevant conclusions