WP4 Construct common model software implementation tool

1
WP4 Construct common model software
implementation tool

• Otto Anker Nielsen, oan_at_ctt.dtu.dk
• Professor, Ph.D.
• Technical University of Denmark (DTU)
• Centre for Traffic and Transport (CTT)

2
Presentation overview

• Introduction to CTT, CLG and Rapidis
• Examples of model software tools
• WP 4 Purposes
• GIS for databases
• Model builder for linking models
• Assignment models
• Work plan and bottlenecks

3
Centre for Traffic and Transport

• One of the 15 centres and departments at DTU
• Department-like structure with teaching and staff
  (like Civil Engineering)
• Established January, 1999 to visualize and
  strengthen the profile of DTU with respect to
  traffic and transport research
• Around 25 employees (3 full professors)
• Rapidis Subcontractor (www.rapidis.com)
• 25 courses (4 shared courses)
• 70 externally financed, 30 from DTU
• 30 external projects, 3 mil Euro over 2-4 years
• www.ctt.dtu.dk

4
Research focus at CTT

• Research in traffic and transport
• Main focus is quantitative methods and modelling
• Transport modelling (Otto Anker Nielsen)
• Transport optimisation (Oli B.G. Madsen)
• Decision support systems (Steen Leleur)
• All modes
• Passenger transport and freight
• All scales (from intersection simulation to
  worldwide air transport)

5
CTT Areas of Interest

• Logistics and Transport
• Traffic and Transport Models
• Geographical Information Systems
• Decision Models and Evaluation Methods
• Traffic Informatics
• Traffic Planning
• Traffic Engineering
• Highway Engineering
• Rail Engineering

6
Strategy

• Coherence between research, development, tests
  and applications
• Research at international level
• Development of effective solution algorithms for
  large scale problems
• Practically useful improvements of methods
• Education based on research activities

7
EU projects

• Project evaluation and decision support
• APAS/ROAD3, EUNET, TEN-ASSESS, CODE-TEN,
  TRANS-TALK and TEN-INVEST
• Transport modelling for decision support
• BRIDGES, SPOTLIGHT, TRANSTOOLS
• Dissemination of EU-research
• PORTAL
• Road pricing
• AKTA/PROGRESS
• Public transport and terminals
• MIMIC
• Air transport optimisation
• DECARTES

8
Courses at CTT and DTU
9
Centre for Logistics and Freight Transport

• Purpose
• Improve research on freight transport, logistics
  and modelling
• Increase knowledge about actors in freight
  transport and logistics
• Development of methods and concepts to be used
  within the sector
• Multidisciplinary co-operation
• Funded by the Danish Technical Research Council
  (STVF)
• Started late 2001, Budget 25 mill. DKr (7 DKK 1
  Euro)
• 9 Danish research institutions
• Participants
• 10 established researchers, 7 research
  assistants, 6 Ph.D. studies, 8 post docs
• www.clgdk.com

10
Rapidis (www.rapidis.com)

• Subcontractor to CTT in TRANS-TOOLS
• Owned by 4 partners, Informi GIS and ESRI
• Specialised in development modelling software
  within the transport sector
• Especially concerning GIS-integration
• Assignment models
• Research project and various projects for transit
  authorities, logistics and network algorithms

11
Examples of models

• Passenger travel model for Copenhagen (OTM),
  1995-
• Harbour tunnel Model
• Passenger travel model for large railway projects
  (KRM), 1999-
• National Danish freight model, 2004-
• AKTA road pricing model
• Strategic model for long-term household decisions
• EU BRIDGES software
• Sketch models in various European projects

12
Experiences with software

• Own implementation
• Multi-modal, multi-class, SUE Traffic assignment
  procedures
• Schedule-based and frequency aggregation
• Passenger and freight
• From local (intersection delay, train simulation,
  access modes and transfer between modes) to
  global (container flow, air passenger flows)
• Standard software
• Passenger models TransCAD, EMME/2, CUBE (TRIPS),
  MOTORS
• Freight models STAN, CUBE Cargo
• Strategic models MEPLAN, SCGE modelling software
• Simulation models VISSIM, Paramics, RailSYS,
  UXSimu

13
Persons involved from Denmark

• CTT
• Otto Anker Nielsen, Project Responsible,
  Professor, Ph.D.
• Christian Overgaard Hansen, Project Leader,
  Associate Professor, Ph.D.
• Christian Würtz, M.Sc. Data issues in GIS, tests
  of models
• Alex Landex, M.Sc. Schedule based transport, GIS,
  tests of models
• Rapidis
• Rasmus Dyhr Frederiksen, M.Sc., Assignment models
• Thomas Israelsen, M.Sc. And Bjarke Brun, M.Sc.
  Model development in GIS, assistance with model
  builder
• TetraPlan
• Possible advises on user interface and freight
  models

14
WP structure overview
15
Description of work to be carried out in WP4

• Construct of input databases
• Scanning possibilities of an open architecture
• Construct of software platform and user interface

16
Why use a GIS-platform?

• Promised in proposal
• Existing framework for user interface and import
  export of data
• Can embed various types of databases and data
  sources
• Can easily join different data either by data
  base linkage but also by coordinates
• Built in procedures to transfer between
  coordinate systems
• Model builder tool box for Geoprocessing
• Can integrate software developed in most
  languages and formats
• Used in Eurostat and DGTREN

17
Background East Denmark Model (CRM)

• A detailed traffic model covering half of Denmark
• Timetables (all runs with about 1000 lines 0.4
  million stops, 30,000 stop groups and 4 million
  pseudo arcs)

18
Background - CRM

• Complex dataflow, Difficult to maintain
  consistency
• Separate models used proprietary data formats

19
Background BRIDGES and SPOTLIGHTS

• EU-sponsored international research program
• Aim
• Better use of GIS in Traffic Models
• Develop exchange format for Transportation data
• Develop data models for complex topological
  objects
• Result Data models designed, but insufficient
  technology
• SPOTLIGHT Refinements of BRIDGES, adoption of
  TOP-objects, European GTF-standard

20
Background - Technology

• Newest generation of GIS is
• Object Oriented
• Flexible - new and customisable data models, etc.
• Extensible Programmable - embedded
  functionality
• Open - Standard DBMS, COM, VBA etc.
• Offers advanced features - tools, versioned
  editing etc.
• It is now possible to create new topological
  objects,
• design new data models and embed functionality!

21
Data model for passenger transport (1)
22
Data model for passenger transport (2)
23
Data model for passenger transport (3)
24
Implementation of model in ArcGIS
ESRI Classes
SimpleJunctionFeature
ESRI Classes
Feature
ESRI Classes
SimpleEdgeFeature
ESRI Classes
Object
ChangeEdge
Connector
DEMAND
Terminator
Turn
TransportEdge
Matrix
MatrixElement
TransportJunction
PHYSICAL
Run
RelationshipClass
ROUTE
NETWORK
StopEdge
RouteSegmentDetail
NETWORK
RelationshipClass
RouteDetail
FrequencyRun
Route
Stop
RouteSegment
StopJunction
RelationshipClass
Terminals
TimePatternDetail
TimePattern
DiscreteRun
25
Example on passenger model

• Short demo

26
Main structure of the GIS model for freight
transport

• 4 Main groups of objects

27
Some examples - Scalability

• Terminal in Copenhagen

28
Zoom out

• Scheduled rail transport

29
Intercontinental level
Long distance lines
30
Global level
31
Zoom to regional level
32
Local mode choice
33
Correspondence to local feeder-line
34
Model Builder

• New tool in ArcGIS 9
• Easy way to define relationships between
  sub-models of any type
• Aggregate model environment for scenarios,
  management of data and models
• Somewhat similar to TRIPs model interface
  although not restricted to specific software for
  data end models

35
Example of coding a 4-step model
36
Trip distributions details within sub-models
37
Integration with GIS standard procedures

• Calculation of hinterlands

38
Join of different data sources

• Transport networks
• Land use
• Transport terminals
• Hinterlands

39
Practical implementation

• GIS model linked to external assignment/route
  choice model

40
Levels of integration of sub-models

• ArcGIS Geoprocessing tool
• The data format is open, and besides the ESRI
  formats, it can use a wide array of other data
  sources, including MS Access, Oracle Database,
  SQL Server Database, Text files etc. The version
  requires ArcView 9 to run
• Most likely some import/export routines and
  processing of results
• As a .NET component for use with C, VB.NET or
  C.NET
• Most likely the assignment models
• As a COM component for use with C or VB 6 (or
  other COM-compliant languages)
• As a stand-alone executable program
• Run by control file
• Input output as text-files, databases or binary
  files
• Most likely the demand models

41
Core issues

• Joint understanding and definitions of variables
• Consistence between LoS in different sub-models
• Pivot-methods between model steps
• Aggregation and dis-aggregation procedures
• Outer loop (equilibrium), e.g. MSA

42
Assignment models

• IPR free existing models or modification of
  various CTT models?
• An equilibrium is obtained where no travellers
  perceived utility
• determined by
• the traveller class utility function
• the travellers preferences
• and the type of vehicle (service) and its
  reliability
• can be increased solely by the traveller changing
  route at the desired time of travel

43
Restrictions

• Utility can be flow-dependent
• Both concave and convex relationships
• Modes can be delayed
• Statistical/empirical distributions
• Simulation models (e.g. rail) gt Not realistic
  within timeframe
• Time-tables preferable, but can be frequency
  aggregated

44
Utility functions

• Typical discrete choice model
• Route choice problem
• Assignment model

45
Why use Error Components

• Different preferences (i.e. weights) of
  attributes compared to each other
• Rather than choices between a few discrete number
  of alternatives

46
Example on arc-based variation
47
Variation of the coefficients

48
Choices in freight transport

• Warehouse at producer (just in time production
  principle)
• Ufunction of transport PC (time, transport cost)
  risk of late delivery possible higher product
  prise (including warehousing cost)
• Intermediate warehouse
• Ufunction of transport PW (time, transport cost)
  transport WC warehousing cost(low due to
  volume) capital cost of storage lower risk of
  late delivery
• Warehouse at the consumer
• Ufunction of transport PC (time, transport cost)
  warehousing cost(maybe higher due to less
  volume and need for longer dwelling time)
  capital cost of storage

49
Choice of mode and route

• 3-level hierarchy
• Decision by transport buyers (choice set
  restrictions)
• E.g. door to door trucks versus multi-modal
  transport chain
• Discrete choice models (standard methods)
• Choices in multi-modal networks
• Terminals and route-segments between terminals
• Discrete choice (choice set reduction),
  simultaneous route and sub-mode choice, or
  combination
• Assignment and route choice along paths
• Routes between terminals, e.g. truck drivers
  route choices

50
Non-linear cost functions
51
Capacity at route segment level

• Not traditional generalised cost / flow curves
• Travel times may reduce with freight volumes
  (higher frequencies, more efficient means of
  transport)
• Travel costs may reduce with freight volumes
  (larger more efficient means of transport)
• Supply may be supposed to fit demand fairly
  efficiently
• In estimation context this may be coded exogenous
• BUT what in the forecast context?
• Can be dealt with by transport logistic modules
• But simple reasoning on generalised
  cost-functions a big step forward

52
Why use random coefficients in freight models?

• Variation of preferences within a certain
  category of goods / freight transport
• In general large heterogeneities within a sector,
  e.g. value of the goods and the preferences by
  decision makers
• Variation of time restrictions
• Different business concepts
• General differences in characteristics
• gt Captures heterogeneities, i.e. not related to
  stochastic behaviour

53
Why use error terms?

• In some circumstances it is a myth that freight
  transport act more rational in term of
  generalised transport cost (utility) than
  passenger transport
• The cost of knowledge acquisition (time use)
• Transport cost infinitesimal compared to
  production cost, capital cost, ware house costs,
  etc.
• Habits (i.e. fixed partner of co-operation)
• Own transport departments (fixed versus variable
  costs)
• All this may be captured by the error term
• Somewhat substitute for lacking explanatory
  variables in the utility function

54
Other reasons for the error term

• Consider the overlapping route problem
• Enables Random Utility Framework and estimation
  of discrete choice models
• Capture unexplained variation (together with the
  alternative specific constant)

55
Recommendations

• Try to replicate existing models
• But may consider to add error terms and EC
• Secure that LoS that is feed-backed is consistent
  with prior model assumptions

56
WP4 Draft working plan and timetable
57
Working plan and bottlenecks

• WP 4 depends heavily on input from WP 2 and 3.
• Time-schedule and budget for WP 4 is very tight
• Programming of the model framework can first be
  initiated when WP 3 have been finalised, or all
  formats have been agreed upon
• gt Short time for implementation
• gt Delays in WP 3 will result in similar delays
  in WP 4.
• Following changes (except marginal) in models
  cannot be accepted due to time and budget
  constraints in WP 4.
• Detailed time schedules should be defined for all
  WPs and the interaction between WPs