DVRPC TMIP Peer Review

TIM 2 ModelOct. 29th, 2014

Travel Models - Overview

• TIM1.0• First VISUM model, completed in 2009

• TIM 2.0• “Best-in-class” 4-step model• Networks carry forward

• TIM 2.1 & TIM 2.2• Minor bug fixes and improvements

• Tim 3.0• Fully disaggregate microsimulated activity

based

Travel Models – TIM 2.0

Classical 4-step model 4 times-of-day 10 trip purpose x income types Limited bike and walk trip model Transit by access mode – walk vs. drive Large network – 26 counties Open source networks – OSM and GTF Iterative feedback for system convergence 12-hours to run, 90 GB disk space, 48+GB

RAM

Model Enhancements

• Larger area• More time periods• More detailed highway network• More Traffic Analysis Zones• Better representation of transit service and

fares• More socioeconomic variables• More trip purposes• Better treatment of non-motorized travel• Improved model operations

Extension of the Model Area Simplified

representation of adjacent Counties

Expected benefits: Full coverage of

DVRPC’s home-work travel shed

Easier start-up for studies across MPO boundaries

The OpenStreetMap www.osm.org

Started in 2004 Organization: OSM foundation

non-profit, based in the U.K. Volunteers

They generate the map Upload data from their private GPS devices Edit directly on www.osm.org

Data distribution Free of charge Can be used for any commercial or non-commercial

purpose Data content

Routable street network plus other geography U.S. data derived from an import of the 2005 TIGER file

GTFS Overview

TriMet/Google developed specification Widely adopted standard for public

transit Series of text files with comma-delimited

values (GTFS = General Transit Feed

Specification)

Open Data Mash-up for Modeling

Data integration Data objects

of different origin are merged

New relationships are created

from OSM

Stop Point

Number

Line

Name

Service Pattern

Line NameRoute NameDirection

Scheduled Run

Line NameRoute NameDirectionIndex

Travel DemandData

Stop Area

Number

from GTFS

Node

Number

Link

From NodeTo Node

2

1 or more

0 or more

Exactly 1

Legend

Connector

Zone NumberNode NumberDirection

Zone

Zone Number

The TIM 2.0 Network in Numbers

Number of network objects TIM1.0 TIM2.0

Street segments 50,000 580,000

Transit stops (stop points) 5,000 18,000

Transit service patterns 2,000 6,000

TAZ (traffic analysis zones) 2,000 3,400

More Detailed Highway Network

TIM1.0

TIM2.0

Integrated Street & Transit Network

Highway Assignment Example

Demand Model

• Conventional 4-step model• Uses “Hotstart” approach –

generic trip table and PnR lot choice pre-loaded for quick convergence

• 12 hours run time• 5/5/5/3 iterations for

AM/MD/PM/NT• Trip purpose segregation by

income • TD & MC done using nested

logit model in VISUM

Trip Generation

Trip Distribution

Mode Choice

Highway Assignment

TIM 2.0 Flow

Trip Purposes

• Home based work• Low income• High income

• Home based shop• Low income• High income

• Home based school• Home based university• Home based other

• Low income• High income

• Non- home based work• Non home based other

Trip Generation

TOD TG

Balancing

Directional Factoring

to TD

Trip Gen Approach:

Ps As

Os Ds

Os Ds

Ps As

• Daily trip generation using rates and demographic variables

• Motorized / non-motorized mode split using logit models

• Time of day factoring• Directional factoring

to Os and Ds

Socioeconomic Variables

• Households by• Size• Number of

workers• Number of autos• Income category

• School Enrollment• K-12• University

• NAICS Employment• Professional services• Eds and Meds• Arts/Rec/Food services• Other services

• Land Use Variables• Parking cost• Retail density• Land use mix• Concentration of low

income households

Trip Distribution & Mode Choice

• Trip distribution in O-D format (e.g. Home to Work handled separately from Work to Home) using gravity model

• Transit trips nested by mode of approach

All trips

Highway Transit

Transit Walk Transit Auto

TIM 2.0 Park and Ride Modeling

Basic Method:

Step 0 – Prepare TIM 2.0 zone system (6 virtual P&R zones)

Step 1 – Obtain highway (ik) and transit walk-access (kj) skims

Step 2 – Matrix Convolution – determine ij skim matrix Step 2.1 Determine Optimal Lot (k) for each ij pair Step 2.2 Compose ij skim from highway portion (ik) and

transit portion (kj) Step 3 – Determine Auto access trip table via nested

logit model Step 4 – Split and assign joint auto-transit trip

Step 4.1 Split ij joint trip into highway and transit portions Step 4.2 Add ik portion to total auto trip table and assign Step 4.3 Add kj portion to transit-walk trip table and

assign

Terminology

i – auto trip end

k – P&R zone

j – transit trip end

Model execution

• Masterbody script handles overall model execution

• Mixture of “canned” VISUM functions and python scripts

• ~90 GB free disk space needed

• At least 16 GB RAM for sequential execution

• At least 64 GB RAM for parallel execution

• Visum will use as much computational power as you can provide (current machines have 12 cores)

Overlord

Calendar Scheduler

MasterScript

(Python)

Daily Trip Gen

AM Model

Midday Model

PM Model

InitialMain Body

Imp. AveragingTrip Dist.Mode ChoiceHwy. Assignment

End

Implementation of TIM 2.0 in VISUM –Model Architecture

Night Model

Reporting Tools

outputs

copies inserts

Visum Run

Report CSVs

VBA + Batch

Scripts

Excel XLSM

Print Collated

Validation Reports