1、Sensitivity Testing of Activity-Based ModelsTRB Innovation in Travel Analysis and PlanningIndianapolis,IndianaJune 5,2023Presented by Andrew Rohne2Introduction Testing of model scenarios during implementation of activity-based models(ABM)Part of ActivitySim Implementation Projects Metropolitan Washi
2、ngton COG(MWCOG)Washington,DC Metropolitan CouncilMinneapolis-St.Paul,Minnesota Southeast Michigan COG(SEMCOG)Detroit,MichiganWhy?Testing fidelity of these models Testing and calibrating sensitivity of the model Understanding the model Model training Testing model features3General Test LogisticsTest
3、 mid-late in calibrationCalibrate sensitivity of the modelTest needs to be specific A-B test“A”is the base scenario“B”is limited changes to that base scenarioDocumentation and discussion is criticalTest methods “what,why,how”Comparison of expected outcomes(vague)and model outcomes(specific)Discussio
4、n of impactIs the model response“in the range of expectation”?Model Status4SEMCOGMWCOGMetropolitan CouncilPlatformTransCAD+ActivitySimCube+ActivitySimCube+ActivitySimStatusModel CompletedSome testing remainsPhase 2(full calibration)nearing completionPhase 2 starting soonFirst ABMFirst ABMReplacement
5、 ABM4.8 million people2,811 TAZ/28,637 MAZ4,600 square mi7.2 million people3,669 zones(internal)6,919 square mi3.6 million people3,030 zones(internal)10,190 square miSensitivity Tests5MWCOGPhase 1Metropolitan CouncilPhase 1SEMCOGBehavioral Telecommute FrequencyAuto Operating CostToll RatesTelecommut
6、e FrequencyAuto Operating CostTNC PricingHousehold IncomeNetworkBridge ClosureTransit FrequencyNew Transit ServiceNew Transit ServiceTelecommute Frequency Sensitivity Tests6RegionMWCOGMetropolitan CouncilTestIncreased telecommute for workers that work in DC by 50%Increased telecommute regionally by