Activity-Based Model For Atlanta (Day 1) by Guy Rousseau, Atlanta Regional Commission Based on the CT-RAMP (Coordinated Travel – Regional Activitybased Modeling Platform) Family of Activity-Based Travel Demand Models • • Main features: • Explicit intra-household interactions • Continuous temporal dimension (Half-hourly time periods) • Integration of location, time-of-day, and mode choice models •Java-based package for AB model implementation Features of CT-RAMP for ARC • Combination of best features of developed AB models: – Fully disaggregate micro-simulation of daily patterns – Consistent treatment of travel tours • Addresses specific planning needs of ARC: – Dynamic / changing population – Toll facilities and managed lanes • Innovations: – – – – Intra-household interactions Toll choice “Available” time influences travel generation Parking location choice Treatment of Space • 2027 TAZs Hall Bartow Cherokee • TAZs subdivided into transit accessibility: Forsyth • Short walk (1/3 mi) • Long walk (2/3 mi) • No walk (> 2/3 mi) Barrow Gwinnett Paulding Cobb Walton DeKalb Douglas • All origins and destinations identified by TAZ and sub-zone Fulton Rockdale Carroll Clayton • 6081 total alternatives in destination choice Newton Henry Fayette Coweta Spalding ¯ ARC ABM Evolution / History • 2001 -> 2002: Household Travel Survey Data Collection & Analysis • 2003 2006 – Models estimated, population synthesizer developed (as presented @ ITM 2006 in Austin TX) • 2007 2008 – Model implementation, calibration started • 2009 2010 – Calibration/validation completed, documentation, deployment at ARC, and sensitivity testing • 2011 – Enhanced data reporting and visualization of outputs 4 Person Types NUMBER PERSON-TYPE AGE WORK STATUS SCHOOL STATUS 1 Full-time worker 18+ Full-time None 2 Part-time worker 18+ Part-time None 3 Non-working adult 18 – 64 Unemployed None 4 Non-working senior 65+ Unemployed None 5 College student 18+ Any College + 6 Driving age student 16-17 Any Pre-college 7 Non-driving student 6 – 16 None Pre-college 8 Pre-school 0-5 None None Activity Types TYPE PURPOSE DESCRIPTION CLASSIFICATION ELIGIBILITY 1 Work Working at regular workplace or work-related activities outside the home. Mandatory Workers and students 2 University College + Mandatory Age 18+ 3 High School Grades 9-12 Mandatory Age 14-17 4 Grade School Grades K-8 Mandatory Age 5-13 5 Escorting Pick-up/drop-off passengers (auto trips only). Maintenance Age 16+ 6 Shopping Shopping away from home. Maintenance 5+ (if joint travel, all persons) 7 Other Maintenance Personal business/services, and medical appointments. Maintenance 5+ (if joint travel, all persons) 8 Social/Recreational Recreation, visiting friends/family. Discretionary 5+ (if joint travel, all persons) 9 Eat Out Eating outside of home. Discretionary 5+ (if joint travel, all persons) 10 Other Discretionary Volunteer work, religious activities. Discretionary 5+ (if joint travel, all persons) Treatment of Time • Time-of-day choice models work on hourly periods • AM and Midday skims used in choice models • Output trips assigned by 5 time periods for highway, 3 for transit NUMBER DESCRIPTION BEGIN TIME END TIME 1 Early 3:00 A.M. 5:59 A.M. 2 A.M. Peak 6:00 A.M. 9:59 A.M. 3 Midday 10:00 A.M. 2:59 P.M. 4 P.M. Peak 3:00 P.M. 6:59 P.M. 5 Evening 7:00 P.M. 2:59 A.M. Treatment of Modes • Explicit toll versus non-toll choice in mode choice • Local versus Premium (express bus, BRT, rail) transit Implementation Design Goals • Overnight run time Model Relevance • • Around 12 to 16 hours Requires distributed processing and threading via Cube Cluster • Commodity hardware Minimize total lifetime cost • Hardware available today from common vendors; reasonably priced • Easy to Setup and Use Staff acceptance • Not too complicated to setup, run, debug, etc ARC ABM Run Times (min) 0 Network Prep, Truck Model, Initial Skims II Demand with CT-Ramp (33% Sample) Convert Trip Lists to Demand Matrices Highway & Transit Assignment & Skimming II Demand with CT-Ramp (50% Sample) Convert Trip Lists to Demand Matrices Highway & Transit Assignment & Skimming 1000 Highway & Transit Assignment & Skimming Highway Assignment (AM, PM, MD, NT) Total 3000 4000 5000 6000 7000 8000 9000 33 25 No Threading/Distribution (8 processors, 16GB RAM, 1 Computer) 1400 112 Threaded and Distributed (24 processors, 48GB RAM, 3 computers) 24 6 170 80 2100 165 36 9 170 75 4200 II Demand with CT-Ramp (100% Sample) Convert Trip Lists to Demand Matrices 2000 310 52 13 173 75 437 100 8795 970 10000 Hardware and Software Setup • Three Windows Server 2003 64bit Machines: • Dual Quad Core Intel Xeon X5570 2.93 GHz with HyperThreading 16 threads • 32 GB of RAM • Cube Voyager + 8 seat Cube Cluster license • Total cost ~ $30,000 in 2009 Hardware and Software Setup • • • • • 64 bit OS for large memory addresses 64 bit Java for CT-RAMP 32 bit Java to integrate with Cube’s native matrix I/O DLL Cube Base for the GUI Cube Voyager + Cluster for running the model, assignment, etc • Java CT-RAMP software • 64 bit R for reporting/visualization Overall System Setup • Cube runs the show and calls all Java processes • User starts the remote processes on the 2nd and 3rd machine (for now) • Everything talks to one mapped network folder location Model Validation Highway Trip-Based Model Year 2005 Volume/Count Scatterplot 200000 180000 Year 2005 Volume/Count Scatterplot 200000 180000 R² = 0.9489 160000 160000 140000 140000 Model Volumes Model Volumes Activity-Based Model 120000 100000 80000 60000 R² = 0.941 120000 100000 80000 60000 40000 40000 20000 20000 0 0 20000400006000080000100000120000140000160000180000200000 Counts X… 0 0 20000 40000 60000 80000100000120000140000160000180000200000 Counts X… New Measures New Measures Persons Not At Home by TAZ New Measures Persons by TAZ New Measures ABM Visualization & Reporting System Activity-Based Model (Java, Cube) Database (SQL Server) Custom Analysis Data Access Layer (IIS, ASP.Net) Visualization Dashboard (Flash) Reports (Excel) ABM VIZ – Time Use • • New time use (person activity over the day) Can select different person types (the above is showing Full-time workers) Radar Charts