TMIPEMAT for Exploratory Analysis presented to presented by
TMIP-EMAT for Exploratory Analysis presented to presented by FSUTMS Cambridge Systematics, Inc. Marty Milkovits November 15, 2019 1
Acknowledgements and Disclaimer Federal Highway Administration Sarah Sun Cambridge Systematics Rachel Copperman Jason Lemp Jeffrey Newman Thomas Rossi The views expressed in this presentation do not necessarily represent the opinions of the FHWA and do not constitute an endorsement, recommendation, or specification by the FHWA. 2
Outline Motivation and Challenges Introduction to TMIP-EMAT Example Application Summary and Next Steps 3
Motivation and Challenges 4
Uncertainties to Scenarios URBANIZATION Plausible Scenarios TECHNOLOGY DEMOGRAPHICS GLOBALIZATION CLIMATE 5
Available Tools Regional travel demand model » Trip-based » Activity-based Land use Outputs Core Model Data driven Inputs, correlations, assumptions Sketch (strategic) 6
Current State of the Practice Uncertainty Space Point prediction Best guess on all concerns Scenario planning Several best guesses We should do more (and we can) Core Model 7
Robust Decision-Making Mitigation Strategies Performs well (no regrets) across plausible futures? STRATEGIES TO MITIGATE IMPACTS OR SHAPE FUTURE No Reliable signposts? Shaping Strategies No Yes Sufficient lead time? No Yes Near-term robust strategy (low risk) 8 Deferred adaptive strategy (low risk) Near-term hedging and shaping strategies (higher risk)
Introduction to TMIP-EMAT 9
What is TMIP-EMAT? EMAT: EXPLORATORY MODELING AND ANALYSIS TOOL Development funded by FHWA Travel Model Improvement Program Tool to support a quantitative Robust Decision-Making approach to transportation planning with deep uncertainty Complements and enhances (does not replace) existing models, visualizations, or planning tools 10 10
TMIP-EMAT Workflow 11 Define the uncertainty and decision space Scoping Run model across uncertainty / decision dimensions Model Risk / Exploratory analysis Analyze
Scope Model Workflow Details Step 1: Scoping—Define uncertainty and decision space Step 2: Meta-model development to produce outcome space Step 3: Simulation (populate outcome space) and analysis Trivial core model runtimes: simulate directly Monte Carlo simulation of experiments § Strategy levers § Metrics § Ex. Uncertainties Non-trivial meta-models Simulate Scoping Analyze Meta-model development Where necessary, leverages Core Model outputs to produce Meta-models that can quickly explore the range of uncertainty 12 § § § Design experiments Run experiments in core model Derive meta-model Meta-models are regression models of the Core Model outputs that run very fast. Metrics by Experiment Risk analysis Exploratory analysis
TMIP-EMAT Components Scoping & Experiment Design Levers Metrics Uncertainties Model Manager & Meta-Model Development Analysis Tools & Visualizers 13 Experiment Database Core Model API Core Model STANDARD TMIP-EMAT COMPONENTS DEPLOYMENT SPECIFIC REQUIREMENTS REGION/APPLICATION SPECIFIC MATERIALS
Core Model API to TMIP-EMAT p y t h o n Create Scenarios Core Model Launch Model Generate Output Summaries A P I Import Outputs Core Model Script 14 Python TMIP-EMAT
Demonstration 15
1 6 TMIP-EMAT Deployments Beta. Tester Model Software Platform ODOT VISUM SANDAG GBNRTC 16 EMME Trans. CAD Model Type Test Application Activity-Based Model Regional analysis of transit, and parking strategies with uncertainty on new vehicle technology Tour-Based Sub. Component Sub-regional analysis of border access highway policies and transit investments with uncertainty around changing demographics, land use, border access, and new vehicle technology Trip-Based Model Corridor-level analysis of transit and complete streetstype improvements with uncertainty on land use, new mobility services, and weather impacts
Visualization Demo with Oregon DOT The purpose of this test and the following analysis was to evaluate ODOT’s new Activity Based Model (ABM); specifically the ability of the ABM to provide information about emerging technologies. To help to achieve that purpose a realistic, but fictitious, set of regional ABM inputs was developed. At the end of this beta test, several flaws in the performance measure creation and methodology were noted as potential improvements for future analysis, but were not corrected in this dataset and resulting analysis. The information in this data and analysis serves as an example for how to use TMIP-EMAT using realistic data. This dataset and analysis should not be used to draw any specific conclusions about transportation policy’s impact on system performance and outcomes. 17
ODOT Beta-Test Scope Policy/Strategy: Uncertainties: » Replace all fixed-route transit with direct service » Generalized transit investment » Parking policy » Micro-mobility » Vehicle technology Performance Measures: » Accessibility measures » Congestion and vehicle usage » Multi-modal usage: mode split, transit ridership » Household activity and mobility 18 § Freeway capacity § Auto operating costs » Demographics (household incomes)
Launch Jupyter Notebook 19
Thank You! Marty Milkovits – mmilkovits@camsys. com Sarah Sun – sarah. sun@dot. gov Open Source Software: https: //github. com/tmip-emat Documentation: https: //tmipemat. github. io/source/emat. intro. html 20
- Slides: 20