FSUTMS-Voyager: Transit Standards within Evolving FSUTMS Technical Presentation

1
FSUTMS-Voyager Transit Standards within Evolving
FSUTMSTechnical Presentation

• Florida Model Task Force
• Tampa, Florida
• December 12th, 2006

85 slides
2
Topics

• Context/Background
• PT 109
• Methodology findings
• FSUTMS-Voyager Transit Model Guidelines
• Path-building/mode choice
• Network/system coding
• Access
• Assignment calibration/validation
• Final thoughts

3

• Context

4
Transit Path-Builders

• Two types single-path multi-path
• Single-path
• Exclusively available in Tranplan, Minutp TP
• Dominant path-builder in Florida US
• Multi-path
• Available in TransCAD Cube-Voyager (PT)
• Becoming more prevalent

5
Multi-Path vs. Single-Path
Rail/Shuttle
Local Bus
O
D
Express Bus
Path Weighted Time Single-Path Skim Values Multi-Path Skim Values
Rail/Shuttle 55 min 100 45
Local Bus 90 min -- 20
Express Bus 65 min -- 35
Also loading percentages!
6
Advantages to Multi-Path Builders

• Reflect sensitivities that would otherwise be
• Impossible in single-path builders, or
• Create inconsistencies between the path-builder
  mode choice model
• More complex ones can combining headways across
  different modes evaluating multiple boarding
  points
• Helpful for complex Bus Rapid Transit (BRT)
  systems
• May curtail magnitude of bias constants
• Offer better consistency between path-builder
  mode choice weights
• No new single-path builder has been offered by
  the software industry in the past 20 years
• Citilabs single-path builders will not receive
  any long-term enhancements

These factors make multi-path builders
potentially attractive!
7
Multi-Path Builder Unknowns(Early 2006)

• How to
• Design a multi-path model
• Coordinate it with the mode choice model
• Calibrate validate
• Introduce a new transit mode
• Whether they meet FTA New/Small Starts guidance
• Whether they work as intended/desired
• No PT models known to exist

8

• Background

9
Work Task Overview

• Develop new transit standards, keeping mindful
  of
• Existing standards
• User planner needs
• Features capabilities of PT Voyager
• New Starts/Small Starts FTA guidance

10
Existing Transit Model

• Consists of many elements/steps, including
• Percent of zonal area within walking distance to
  transit
• Walk, auto sidewalk connections to transit
• Transit line coding, fare definition, speed
  relationships
• Path-building
• Mode choice modeling
• Assignment
• Reporting
• Each element has to be consistent with the others
  to produce viable results
• Path-builder drives transit model structure!

11
Different Needs
Tier Transit Service Data Availability Examples
A Local service only no park-and-rides System-wide boardings transfer rates Polk County, SunTran
B Local express service some park-and-rides System-wide boardings transfer rates May have recent on-board survey Votran, Spacecoast
C Local express service some park-and-rides Recent on-board survey system-wide boardings transfer rates Jacksonville, Orlando, Tampa
D Many different types of service extensive park-and-ride system potential major fixed-guideway system Recent on-board survey for all modes detailed boarding transfer rate information Southeast Florida
12
FTA New/Small Starts

• Since 2002, FTA found that many ideas considered
  good practice may have many bad or undesirable
  properties during forecasting
• FTA has released recommended model properties and
  other findings to the modeling community in the
  hopes that future modeling systems will avoid
  these practices (see next slide)
• Instituting quality-control tests, two of which
  are impacted by multi-path path-builders
• Overall, FTA struggling with the impact of
  multi-path builders on its evaluation of New
  Starts projects

13
Problematic Characteristics of Transit
Forecasting Methods

• Unusual coefficients in mode choice models
• Non-logit decision rules
• Bizarre alternative-specific constants
• Path / mode-choice inconsistencies
• Accuracy of bus running times
• Stability of highway-assignment results
• Assertions for new transit access modes

14

• PT 109

15
PT Public Transport (1 of 5)

• Public transportation module in Cube-Voyager
• Methodology features very different from
  Tranplan (more later)
• Takes advantage of Voyagers features
• Highlights
• Multi-path builder
• Includes ability to build access connectors
• More powerful line coding
• Flexible auto-transit speed relationships

16
PT Public Transport (2 of 5)

• Three major characteristics different from
  existing FSUTMS transit model
• Multi-path algorithm (vs. single-path)
• Leg-based (vs. link-based)
• Runs in 4 steps (vs. straightforward)

17
PT Public Transport (3 of 5)

• Three major characteristics different from
  existing FSUTMS transit model
• Multi-path algorithm (vs. single-path)
• Impacts path-builder/mode choice relationship
• Leg-based (vs. link-based)
• Runs in 4 steps (vs. straightforward)

18
Leg vs. Link
Centroid
Station
Rail Line
19
Transit Link Example
Centroid
Walk-access connector
Sidewalk link
Station
Rail Line
Sidewalk link
20
Transit Leg

• Two requirements
• All connectors must connect transit stop to
  transit stop or transit stop to centroid
• Connectors must spider highway network

Centroid
Walk-access leg
Station
Rail Line
21
PT Public Transport (4 of 5)

• Three major characteristics different from
  existing FSUTMS transit model
• Multi-path algorithm (vs. single-path)
• Impacts path-builder/mode choice relationship
• Leg-based (vs. link-based)
• Impacts access connectors percent walks
• Runs in 4 steps (vs. straightforward)

22
PT Path-Builder Logic Overview

• 4 steps
• Network simplification
• Minimizes solution set maximize running time
• Minimum Cost Path/AON
• Determines minimum path
• Enumeration
• Determines if other paths are acceptable of
  proceeding to evaluation step
• Evaluation
• Determine weights/percentages of remaining paths
• Network coding is very important to ensure that
  paths progress through these steps as intended

23
PT Public Transport (5 of 5)

• Three major characteristics different from
  existing FSUTMS transit model
• Multi-path algorithm (vs. single-path)
• Impacts path-builder/mode choice relationship
• Leg-based (vs. link-based)
• Impacts access connectors percent walks
• Runs in 4 steps (vs. straightforward)
• Impacts network coding

24
FTA Reactions to PT

• Four meetings with FTA, one including Citilabs
• Strongly recommended micro-coding fixed-guideway,
  park-and-ride stations to better represent
  transfer time
• Existing FSUTMS standards utilize same node for
  rail bus stations
• Confirmed that multi-path builder is not
  compatible with existing FSUTMS mode choice
  structure
• PT v4.0 cannot provide the necessary information
  for New Starts quality control tests

25
Recent Events

• Citilabs added a best-path switch to PT in
  summer 2006
• Addresses FTAs New Starts quality control tests
• Mimics single-path builder from Tranplan
• Not compatible with all parameters/keywords
• Initiated testing different model setups
• To provide empirical data to assist with
  FTA/Citilabs discussions
• Help determine best design for a PT-based transit
  model
• Identify any software or design-related issues
  early on

26

• Methodology findings

27
Transit Model Setups
Setup Access Network Path/ Skims Mode Choice Assignment
PT Multi-path PT PT PT, path for each access mode only Access-only Walk/PNR/ KNR PT
PT Best-path PT PT PT, path for each access transit mode combination Access mode Walk/PNR/ KNR by bus/project/ fixed-guideway PT
PT-TRNBUILD PT, converted to TRNBUILD within model PT, converted to TRNBUILD within model TRNBUILD, path for each access transit mode combination Access mode Walk/PNR/ KNR by bus/project/ fixed-guideway TRNBUILD
- long-term recommendation - short-term
recommendation
28
TRNBUILD

• Single-path builder from TP available in
  Voyager
• Link-based, many similarities to Tranplans
  path-builder
• Used nationwide
• Atlanta, Washington DC, Columbus OH, Northern NJ
• Well-known to FTA
• Reliable history that dates back to MINUTP
• Only being used in Florida for SERPM6
• Future development not supported by Citilabs

29
Transit Model Setups

• All setups coded in Olympus
• Networks connectors generated by PT
• Using XCHOICE in MATRIX
• Creates Summit user benefit files
• Developed a BRT/New Start example to test path
  user benefit impacts against changes in
• Mode numbers
• Travel time
• Station micro-coding

30
Findings

• Developed PT Best-path setup that mimics
  PT-TRNBUILD results
• Multi-path model design results very different
  from Best-path TRNBUILD
• Further research needed to define multi-path
  model setup
• Recommendations
• Short-term proceed with PT Best-path setup
• Separate guidelines for General New/Small
  Starts use
• Long-term track evolving FTA guidance, industry
  progress PT updates goal - evolve to multi-path

31

• FSUTMS-Voyager Transit Model Guidelines
• Must use version 4.1 shipped to you last week!

32
Summary of Changes
Module Change(s)
HNET Coding transit network elements in transportation network
DISTRIB Review trip distribution for validation examine highway speeds from first assignment
TNET Advanced line coding features new mode definitions New system data files with reduced emphasis on ASCII files New auto-bus speed relationships
TPATH New access procedures programs Reduced number of paths in some areas
MODE New coefficients new mode choice structure for some areas New percent walk guidelines
TASSIGN New assignment procedure reporting program
Calibration/ Validation Guidelines!
33

• Path-building/Mode choice
• TPATH/MODE

34
Path-Building/Mode Choice (1 of 4)

• Auto nesting same as existing, larger area models

35
Path-Building/Mode Choice (2 of 4)

• Tier A Areas
• Local service only
• No park-and-rides
• Limited data available
• Build walk-transit drive-transit paths per
  period
• Paths include all modes

36
Path-Building/Mode Choice (3 of 4)

• Tier B/C Areas
• Local express service
• At least some park-and-rides
• At least system-wide boardings, but on-board
  survey likely
• May be planning for New/Small Start project in
  near future
• Build 4 paths per period
• Walk-bus
• Walk-project/premium
• Auto-bus
• Auto-project/premium

37
Mode Choice StructureTier B/C Areas
38
Path-Building/Mode Choice (4 of 4)

• SERPM (Tier D Areas)
• Many types of service
• Extensive park-and-ride system
• Detailed boarding rider data available
• Can be planning for New/Small Start project in
  near future
• Build 8 paths per period
• Walk-bus
• Walk-project/premium
• Walk-MetroRail
• Walk-TriRail
• Auto-bus
• Auto-project/premium
• Auto-MetroRail
• Auto-TriRail

39
Mode Choice StructureTier D Areas
40
Mode Choice Utility Coefficients
Variable Units HBW HBO NHB
IVTT Min -0.0250 -0.0125 -0.0250
IVTT for CR Min -0.0200 -0.0100 -0.0200
OVT (walk- wait-time) Min -0.0500 -0.0250 -0.0500
Fare, parking cost, AOC Cents -0.0025 -0.0025 -0.0050
Drive-access time Min -0.0375 -0.01875 -0.0375
Number of Transfers -- -0.1250 -0.0625 -0.1250
41
Mode Choice Utility CoefficientsRelation to IVTT
Variable Units HBW HBO NHB
IVTT Min 1.0x 1.0x 1.0x
IVTT for CR Min 0.8x 0.8x 0.8x
OVT (walk- wait-time) Min 2.0x 2.0x 2.0x
Value of time /hr 6.00 3.00 3.00
Drive-access time Min 1.5x 1.5x 1.5x
Number of Transfers Min 5.0x 5.0x 5.0x
Path-builder weights equivalent to mode choice
variables weighted to IVTT
42
All-Walk Paths

• Need to compare transit with all-walk path
• Maintain consistency with Tranplan/TRNBUILD
• Prevent overstatement of user benefits
• Can do this in
• Path-builder requires zone-to-zone walk
  connectors
• Mode choice requires all-walk skim
• Need to determine most efficient way

43

• Network/system coding
• TNET/HNET

44
Transit LinesPT Capabilities

• Structure similar to INET
• No line numbers or reference lines!
• Stops positive non-stops negative
• Multiple headways per line ? single file
• Flexible in-line coding layovers, access/
  egress-only stops, circulator coding
• Allows double stops complex routings

45
Coding complex routing
46
Transit LinesTransportation Network

• Three elements should be represented in
  transportation network
• Transit-only links
• Micro-coded station
• Station data
• Transit-only links
• Similar to existing practice, just coding on
  transportation network
• Coded with facility type 69 with special fields

47
Transit LinesTransit-Only Link Fields
Field Modes Description
TBSDIST Bus/mixed-flow Distance (miles)
TBSTIME Time (minutes)
TBSSPEED Speed (mph)
TFGDIST Fixed-guideway Distance (miles)
TFGTIME Time (minutes)
TFGSPEED Speed (mph)
TFGMODE Mode
48
Station Micro-CodingGeneral Fixed-guideway
49
Station Micro-CodingNew/Small Starts
Fixed-guideway
50
Station Micro-CodingNew/Small Starts Bus PNR
51
Station Data Fields
Field Description Default Values
TSNAME Station name --
TSTYPE Type of access 0 not used1 used
FAREZONE Fare zone for zone-based fares Coded on station nodes only
TSRANGE Maximum roadway distance allowed for auto-access connector Typically 2.0 10.0
TSPARK Number of parking spaces --
TSCOSTAM Parking cost in peak period --
TSCOSTMD Parking cost in off-peak period --
TSPNRTERM PNR terminal time 2.0
TSKNRTERM KNR terminal time 0.5
52
System Data

• Defines modes, operators wait-curves
• Transit operator
• Assigned to specific transit lines
• Should be defined by fare policy
• Wait curves
• PT allows curvilinear actual-perceived wait time
  relationships
• Existing method applies ½ headway rule with
  30-minute maximum
• Apply piecewise function for rail modes to avoid
  excessive headway impacts on ridership user
  benefits, for instance
• If headway 15 min, set new headway original
  headway
• If headway is between 15 30 min, set new
  headway 2 (7.5 (headway-15)/4)
• If headway is over 30 min, set new headway 2
  (11.25 (headway-30)/8)

53
Transit ModesExisting Structure

• Modal definitions based on service type only
• Problems
• Local/express definition not favored by FTA
• New bus services (e.g., limited-stop, BRT) dont
  fit into hard categories
• If using express bus mode, extensive workarounds
  needed to properly model speeds biases for new
  bus services
• Forecasting requirements for regional models
  starting to exceed 8-mode definition 30-mode
  software limits
• Propose new structure to
• Take advantage of PTs expanded mode limits
• Prepare for eventual migration to multi-path

54
Modal Definitions (1 of 2)
Number Mode
1 Walk access/egress (centroid-to-stop vice-versa)
2 Auto access
3-10 Other access connectors (for future uses)
11 Fixed-guideway platform to street connectors PNR to fixed-guideway/street
12 Transfer connectors (sidewalks)
13-20 Other non-centroid connectors (for future uses)
55
Modal Definitions (2 of 2)
Number Mode
21 Local express bus
22 Bus rapid transit/premium bus
23 Circulator (e.g., Metromover, Streetcar, Trolley)
24 Heavy rail transit (e.g., Metrorail)
25 Commuter rail (e.g., TriRail)
26 Other mode
27 Project mode (for planning studies)
31-37 Same as 21-27, but for different operator (i.e., county)
41-47, 51-57, etc. As needed
56
Auto-Transit Speed RelationshipsRecent Events

• Transition to PT
• Allows expanded mode definitions unique
  auto-transit speed relationships for each mode
• Data collection
• Tampa (2003) Jacksonville (2005)
• Both surveys show that transit speeds are 70 of
  auto speed in all but a few cases
• No data on limited-stop bus or BRT services

57
Auto-Transit Speed Relationships

• Good time to re-evaluate auto-transit speed
  relationships
• Three possible options
• Maintain piecewise relationship
• Linear or curvilinear relationship
• Linear/dwell time hybrid relationship
• Will review options after transit model framework
  is finalized
• Impact with time of day models?

58

• Access
• TNET

59
Zonal Access

• Compute proportion of zonal area within walking
  distance to transit
• Purpose
• Avoid over-estimating transit trips in large
  (gt1mi2) zones while minimizing required number of
  paths/skims
• Existing standards
• Large areas computed short long-walk coverage
• Small areas did not use percent walks
• Developed using GIS or PCWALK program
• Assumed ubiquitous access from inside zone to
  edge (agreed with access program logic)

60
Zonal Access

• Need to continue practice (Tiers C D only)
• Zone sizes in some models are still large
• Proposed standards
• Use a single ½ mile can walk/cannot walk buffer
• Reduces number of access categories to 3 from 7
• Compute via GIS, maintaining ubiquitous access
  assumption
• Does not agree with walk-connector logic, so
  connectors will have to be reviewed compared to
  percent walk values (more later)
• Recommend review of zone sizes for all models

61
Example of Large Zones
Zone 503 (3 mi2) 2000 Population 7,700
2025 Population 25,000 2000 Employment 2,600
2025 Employment 3,300
503
Zone 1129 (3 mi2) 2000 Population 13,900
2025 Population 18,700 2000 Employment
2,100 2025 Employment 2,300
1129
62
503
1129
63
Walk AccessExisting Method

• Primary connections from centroids to bus stops
• Relies heavily on percent walk calculations
  classifications
• If percent walk is non-zero and no links from
  above, program swept surrounding nodes
• CODW times computed so that separate short walk
  long walk times could be computed inside modal
  choice model

64
Walk AccessProblems

• Percent walks not very good
• Often computed across canals and other barriers
• Although barriers were available in logic, they
  were seldom used
• Sweeps included many questionable connectors
• Detailed examination of maps and aerials showed
  PT procedure often much better

65
Walk Access Inconsistency Example (1 of 2)
Zone 32 53 short walk 100 long walk
Only access from zone well-represented by single
centroid connector, requiring very long walk to
node 6084
66

• Extensive transit service to east of zone
  separated from zone by railroad with no crossings

67
Walk Access Inconsistency Example (2 of 2)

• Calculation driven by bus along arterial south of
  centroid (node 15234 etc.) but blocked from TAZ
  by major canal
• TAZ functionally an island with canals on all
  sides and only access via bridge to node 15238

Zone 1694 92 short walk 100 long walk
68
1694
69
Walk AccessNew Approach (1 of 4)

• Connector Data
• Walk connectors now from PTs GENERATE (next
  slide)
• Adjust percent walks globally until better
  approach can be found
• Retain CODW procedures to minimize impact on
  modal choice model
• Modifying the connector data
• Special-purpose program/script to adjust
  connectors and/or percent walks
• Connector/CODW adjustments on slide after next

70
New Approach (2 of 4)Connector Types
Type Maximum Length Description/Rationale
Centroid-to-stop 1.1 miles Standard walk connectors
Station-to-centroid 3.0 miles Length set artificially high to avoid disconnects between alternatives Apply spline function to over-weight walks over 0.5 miles to avoid excessive walks
71
New Approach (3 of 4)Centroid-to-stop Connector
Modifications
Percent walk Transit stop _at_ centroid connector? Action/Notes
100 Yes No modification to access connectors
100 No Reset percent walk to 0
20 x 100 Yes Reset length of centroid portion to ½ mile
20 x 100 No Reset percent walk to 0
x 20 Yes Delete all access connectors (transit likely not really that accessible at all) Reset percent walk to 0
x 20 No Reset percent walk to 0
72
New Approach (4 of 4)Station-to-centroid
Modifications
Connector Length (miles) Modification Process Maximum Modified Length (miles)
0 x 0.5 No modification to connector 0.5
0.5 x 1.0 No modification to connector for first ½ mile Reset any additional length to (x-½)2 Re-compute walking time 1.5
1.0 x 2.0 No modification to connector for first ½ mile Reset next ½ mile to (x-½)2 Reset any additional length to (x-1)3 Re-compute walking time 4.5
2.0 x 3.0 No modification to connector for first ½ mile Reset next ½ mile to (x-½)2 Reset next 1 mile to (x-1)3 Reset any additional length to (x-2)4 Re-compute walking time 8.5
73
Drive Access

• Existing method
• Used AUTOCON logic to create drive-transit
  connectors between centroids park-and-rides
• Avoided backtracking excessive drive time
• Had problem of disconnecting some zones between
  alternatives, slightly impacting user benefits
• Proposed approach
• Favor scripting AUTOCON logic over using PTs
  GENERATE keyword (see next slide)
• Adding function to exaggerate time on connectors
  beyond range
• Modifications to connectors needed (see slide
  after next)

74
Drive Access Connector Comparison
PTs GENERATE generates circular catchment areas
AUTOCONs logic applies backtracking logic
75
Auto Access ConnectorsNew Method

• Embed station costs to PNR- KNR-transit
  connectors
• Driving time 1.5x
• Terminal time 2.0x
• Station parking cost 6/hr (pk) 3/hr (op)
• Auto operating cost 6/hr (pk) 3/hr (op)
• Auto access connectors passed to mode choice as
  IVTT since already weighted
• Separate connectors needed to bus rail
  platforms
• Why? PT does not allow consecutive non-transit
  legs

76
Sidewalk/Transfer ConnectorsExisting Method

• Allows street-walking to transfer between
  different transit lines
• SIDECON produces sidewalk links around transit
  stations
• Sidewalks coded as INET routes to reflect
  walk-able areas (e.g., CBDs)
• Both SIDECON sidewalks are link-based not
  readily-compatible with PT

77
Sidewalk/Transfer AccessNew Method
Type Maximum Length Description/Rationale
Bus stop-to-fixed-guideway platforms 0.6 miles Needed for station micro-coding
Fixed-guideway platforms to nearby bus stops 0.6 miles Allow for movements between fixed-guideway platforms buses down the street
Bus stop-to-bus stop in CBD areas 0.6 miles Replaces need for INET sidewalks
All three can be generated using PTs GENERATE
keyword
78

• Assignment Calibration/Validation
• TASSIGN

79
Transit Assignment

• Assign each transit path
• Decimalized boardings
• Output is DBF format
• Need to develop program/script to
• Concatenate path loadings
• Report results in a user-friendly format

80
Calibration/Validation Steps

• General
• Calibrate mode choice model
• Validate boardings by mode and/or operator
• 20 within each category
• New Starts
• General steps
• Review trip distribution
• Calibrate end-to-end travel times

81

• Final thoughts

82
Time of Day

• FSUTMS uses auto speeds from 24-hour assignment
  as for HBW mode choice
• HBO NHB use free-flow speeds
• SERPM6 uses time of day
• Peak period trips use AM congested speeds
• Off-peak period trips use free-flow speeds
• Standard FSUTMS design expected for most models

Standard FSUTMS Models Standard FSUTMS Models
Purpose Period Auto Impedances
HBW all Congested
HBO all Free-flow
NHB all Free-flow
Time-of-day Models Time-of-day Models
Purpose Period Auto Impedances
HBW peak HBO peak NHB peak AM congested
HBW off-peak HBO off-peak NHB off-peak Free-flow
83
Highway Modeling Impacts

• Standard highway model practice has strong
  impacts on transit modeling
• SPDCAP table adjustments
• Distribution corrections (e.g., k-factors, etc.)
• 0-car household impedances
• New transit model stresses need for
• Refraining from strong SPDCAP table modifications
• Validating trip distribution
• Also
• Code transit-only links transit station data on
  transportation network
• Stability of highway-assignment results

84
Status / Next Steps

• Developed framework documents to summarize key
  points
• Theoretical coordination of individual parts
• Application parameter settings, technical
  details
• Both available on www.fsutmsonline.net
• FDOT gathering feedback
• Will continue to finalize details procedures
• Need for more speed/delay studies and local
  transit on-board studies
• Transit model training workshop June 4-7, 2007

85

• Thank you!