470 likes | 754 Views
SCAG ACTIVITY-BASED TRAVEL DEMAND MODEL ( SimAGENT ). Presentation at the SCAG Modeling Task Force July 22, 2009. Kostas Goulias University of California Santa Barbara. Chandra Bhat The University of Texas Austin. Ram Pendyala Arizona State University Tempe.
E N D
SCAG ACTIVITY-BASED TRAVEL DEMAND MODEL (SimAGENT) Presentation at the SCAG Modeling Task Force July 22, 2009 Kostas Goulias University of California Santa Barbara Chandra Bhat The University of Texas Austin Ram Pendyala Arizona State University Tempe
SimAGENT for SCAGSimulator of Activities, Greenhouse Emissions, Networks, and Travel • Introduction & Definitions • Examples Policy Analysis Needs • Simagent Phase 1 • Simagent Phase 2 • Project Tasks and Management
Actual Schedule of 3 Persons in the Same Household Male head Female head Child
Conventional (zonal) Models (spatial structure representation) Traffic analysis zones, centroids, centroid connectors, higher functional class network
Conventional (zonal) Models(travel behavior representation) Home Based Work Home Based Shop Home Based Other Non Home Based
The 4-step Model Convert real world into Traffic Analysis Zones – Then convert highways and traffic analysis zones into a set of nodes and links building a graph
Improved 4-step From Rossi Seminar
A Person’s Daily Travel Pattern (conventional model) Shop Home Based Shop by car Non Home Based by car Home Diner Home Based Other by car Home Based Work by Bus • TRIPS: • 2 HBW • 1 HBS • 1 HBO • 1 NHB Work centroid
A Person’s Daily Travel Pattern (activity based model) 7:00 PM Shop Home to Shop by car 5:30 PM Shop to Diner by car Home 9:30 PM 5:15 PM Diner 7:30 PM Diner to Home by car 7:30 AM 9:00 PM Home to Work by Bus Work to Home by Bus 4:30 PM • TRIPS: • 2 HBW • 1 HBS • 1 HBO • 1 NHB • 2 Home based tours (chains) • Timing of all trips • Duration of activity at each location Work 8:15 PM Centroid & XY coordinates
Two Other Household Members Travel Pattern (activity based model) 7:00 PM Shop Home to Shop by car 5:30 PM Shop to Diner by car Home 9:30 PM 5:15 PM Diner 7:30 PM Home to School by Car 9:00 PM Diner to Home by car 3:30 PM 2:30 PM Work to School by Car School to Home by Car School Work School to Work by Car 8:00 AM 9:00 AM Centroid & XY coordinates
All Household Members’ Travel Pattern (activity based model) 7:00 PM 7:00 PM Shop Shop Home to Shop by car Home to Shop by car 5:30 PM 5:30 PM Shop to Diner by car Shop to Diner by car Home Home 9:30 PM 9:30 PM 5:15 PM 5:15 PM Diner Diner 7:30 PM 7:30 PM Home to School by Car 7:30 AM 9:00 PM 9:00 PM Diner to Home by car Home to Work by Bus Work to Home by Bus 3:30 PM 2:30 PM 4:30 PM Work to School by Car School to Home by Car School Work Work School to Work by Car 8:15 PM 8:00 AM 9:00 AM Centroid & XY coordinates
On the zonal system (activity based model) 7:00 PM 7:00 PM Shop Shop 5:30 PM 5:30 PM Home to Shop by car Shop to Diner by car Shop to Diner by car Home Home 9:30 PM 9:30 PM 5:15 PM 5:15 PM Diner Diner 7:30 PM 7:30 PM Home to School by Car Diner to Home by car 7:30 AM 9:00 PM 9:00 PM Home to Work by Bus Work to Home by Bus 3:30 PM 2:30 PM 4:30 PM Work to School by Car School to Home by Car School Work Work School to Work by Car 8:15 PM 8:00 AM 9:00 AM
Some Key Aspects of Activity Based Models • Trips are linked for each person in a day • Timing and durations are included • Entire daily travel patterns are linked • Car use is associated to needs (take child to school, drive together to shop & dine and back ) • Removing a trip or activity has a domino effect on everything else that is not “fixed”
9:30pm 9:00pm Shopping Eating at restaurant 5:30pm 7:30pm 4:00pm 4:30pm 3:30pm 2:30pm 7:40am 9:00am 8:00am Home Work Work School OBJECTIVE: Recreate these patterns for the entire SCAG population and have models to build policy scenarios
Examples of policy analysis needs Source: variety of papers on the four step and activity models
Congestion Pricing • Toll strategies/pricing • Impose a toll and predict elasticity of demand (-0.1 to -0.4) • Conventional models • Predict shifts in departure & arrival time • Observed elasticity lower than predicted • Why? • Time offset (freeing capacity taken by others) • Value of time very different among segments • Entire activity-travel schedule modified by pricing • Activity-based models could address these issues • Predict who reacts to policy at the individual level • Predict activity scheduling and task allocation changes within households
HOV/HOT • Conventional models • HOV as a mode (time and cost) • Overestimate the number of users • The problem is lack of accounting for intra-household interactions and carpool formation • Activity-based models • Include hh-member interactions • Include a car assignment to person model/routine
Parking • Conventional models • Parking duration not modeled • Parking lot = destination of trip • Summary demand by period of day • Activity based models • Explicit estimation of parking duration • Operate at fine temporal resolutions • Can keep track of cars in households Note: more inventory data needed!
Transit fare • Conventional models • Zone to zone base fares • Examine changes in ridership and correlate with fare changes • Activity based models • Transit paths can be developed • The impact of waiting times and costs examined in terms of overall change in scheduling • Too much work? Should we calibrate this to potential for change?
Shorter days and weeks • Conventional models • Not sensitive to work duration • Impose change in trip generation and see what happens • Activity-based models • Activities, travel, and duration of activities are tied together • Changes in work duration and days of the week are explicitly modeled (increases in after work periods, available extra day to do other things and so forth)
Demographic shifts • Conventional models • Very few segments • Operate at OD level • Activity based models • Operate at the individual and household levels • Include full-time vs. part time workers • Include children by age groups • Include many additional segmentations because of synthetic population generation • Key to this region ethnicity!
Car ownership and type • Conventional models • Absent • Number of cars per household • Activity-based models • Explicit car ownership and assignment to persons • Type can be incorporated (including fuel type)
Emissions inventory • Conventional models • Vehicle activity is handled by post-processing • Does not account for within household vehicle assignment and does not produce a vehicle trace -> loss of vehicle use profiles • Activity-based models • Details about who uses each vehicle and when/where • Some produce traces of vehicles during the day • New generation emissions models may be more compatible (?!) with this approach • Some explore dynamic traffic assignment but not final word yet!
Land use & development • Conventional models • Build scenarios and data fed into 4-step • Zone to zone travel time and costs (accessibility indicators) used in land use • Can be done in an feedback fashion for lagged time • Activity based models • Offer opportunity for true integration • Land use driven by location desires (and developer desires) • Travel models use more detailed land use data • We hope for parcel detail but many issues are not solved yet
SimAGENT Vision • Comply with the California Transportation Commission (CTC) 2008 guidelines for RTPs • Create an activity-based model that can address wide range of policies, including: • Economic analysis: location-based welfare, wages, and exports • Equity analysis: change in welfare by household income class • Evaluate the energy use and GHGs produced by households and workers in building space • Comprehensively evaluate economic development impacts • Evaluate time-of-day roadway tolls
Phase 1: Adapt CEMDAP-DFW to SCAG SimAGENT Land Use Design and Forecasting (including demographics) Networks & Attributes Parcels/Zones & Attributes Population in Zones (centroids) Accessibility (aggregate) Population Synthesis External Trips Person Daily Tours-stops & trips Long Term Choices REPORT Airports & Ports Daily Allocation Daily Scheduling All Other (commercial) Passenger & Highway and Transit Post Processor Origin – Destination Trip Interchange Matrices EMFAC Network Assignment
Phase 1: Adapt CEMDAP-DFW to SCAG SimAGENT Land Use Design and Forecasting (including demographics) ADAPTED CEMDAP MODEL Networks & Attributes Parcels/Zones & Attributes Population in Zones (centroids) Accessibility (aggregate) Population Synthesis External Trips Person Daily Tours-stops & trips Long Term Choices REPORT Airports & Ports Daily Allocation Daily Scheduling All Other (commercial) Passenger & Highway and Transit Post Processor Origin – Destination Trip Interchange Matrices EMFAC Network Assignment
Forecast Year Outputs Dynamic Traffic Assignment (DTA) Aggregate socio-demographics (base year) Synthetic population generator (SPG) Link volumes and speeds Activity-travel environment characteristics (base year) Socio-economics, land-use and transportation system characteristics simulator (CEMSELTS) Detailed individual-level socio-demographics (base year) Individual activity-travel patterns Policy actions Model parameters Activity-travel simulator (CEMDAP) Socio-demographics and activity-travel environment Base Year Inputs CEMUS SimAGENT Based on CEMUS (Developed at University of Texas at Austin)
Phase 1 Comparison of Model Scale SCAG 2003 Validated Model • 4192 zones (4,109 internal + 40 cordon, 12 airport, 31 port) • Used also for air quality and GHG (CO2) emission estimation with EMFAC • Highway network includes freeway system (mixed-flow lane, auxiliary lane, HOV lane, toll lane, truck lane, etc.), arterials, major collectors, and some minor collectors • AM peak period (6:00 AM to 9:00 AM) • PM peak period (3:00 PM to 7:00 PM) • Mid day period (9:00 AM to 3:00 PM) • Night period (7:00 PM to 6:00 AM) • No Dynamic Traffic Assignment • Traditional feed forward land use and assignment Dallas – Fort Worth CEMDAP Study • 4,874 Zones (4,813 Internal + 61 External), 18,566 network nodes • 22,185 roadway links (26,799 lane miles) + 9,600 zone connector links • 63 HOV links (37 lane miles) • Highway network used with CEMDAP includes freeways, HOV lanes, major arterials, minor arterials, collectors, ramps, frontages, etc. • Morning off peak (3:00 AM to 6:29 AM) • AM peak (6:30 AM to 8:59 AM) • Mid day off peak (9:00 AM to 3:59 PM) • PM peak (4:00 PM to 6:29 PM) • Evening off peak (6:30 PM to 2:59 AM) • Also tested Dynamic Traffic Assignment • Includes key aspects of the integrated model
PHASE 2: Development of Advanced Version of SimAGENT (2011) • Increase spatial detail • Accessibility reflected in major interactions • Expected to have: • Sensitivity to an expanded repertoire of policies • Integrated land use influences on travel behavior • Enhanced feedback among model components • Enhanced reflection of behavioral interactions • Integrated interfaces with land use, traffic assignment, and EMFAC and/or MOVES
Back to the example of two employed and a child. Home accessibility Work accessibility
Home accessibility Work accessibility
Home accessibility School accessibility
Tasks-Schedule-Budget Training is included in Task 2 also. Tasks 2 and 3 contain extensive testing and validation exercises as well as comparisons with the trip based models