Skip to content

Meeting Notes

Jump to bottom Edit New page
Ben Stabler edited this page Oct 22, 2021 · 36 revisions

10/22/21

  • Reviewed updated visualizer
  • TODs looking better but still an issue with work tour TOD
  • Now tracing work scheduling model for clues
  • Setup transit skims by income class
  • Walk transit mode shares too high, need to trace
  • Will share updated visualizer once issues fixed
  • Need to work on DOS setup, integrate EMME Python, use new skims from Nagendra

10/06/21

  • Continuing working on fixing issues
  • Fixing time periods issues inherited from MTC setup
  • Transit park and ride lot costs in there now
  • Working on three income classes for TVPB in activitysim; EMME setup will come later
  • Bike and walk mode issues fixed and testing now
  • Writing trip matrices updated
  • CDAP for preschool updated; will share updated visualizer
  • Still to do
    • DOS / EMME model runner setup
    • School bus trips being produced?
    • Trip mode issues - no KNR trips on KNR tours; review exclusions in expression files
    • Make sure consistent with survey processing
  • Testing with 100k HHs; will update visualizer with 25% sample
  • Nic working on updated streets file
  • Transit assignment path building weights will need to be adjusted so we have three versions, one for each income bin
    • @Nagendra to update this
    • @Craig share VOTs to use from CS model update
    • @Andrew update Asim so three sets of skims coming in and three trip tables going out
  • Post project, CMAP may want a more comprehensive two full day in-person type training
  • Next call October 22nd

09/24/21

  • Working through issues in model, fixing, and then re-running and comparing to the survey with the visualizer
  • Mostly fixed time period distributions, but still investigating
  • Issues identified with the visualizer comparisons:
    • Preschool person types making mandatory tours?
    • No trips after 4pm?
    • Bike and taxi mode way too high and SOV low
      • Maybe intrazonal distances are 0 which doesn't mean NA but 0 impedance (i.e. travel at lightspeed)
      • Check expressions and input files, especially MAZ to MAZ files
    • No school bus trips
  • Other todos
    • Need to clean up output matrices for assignment
    • Add parking cost at taps to maz tap file and expression
    • Go from 2 sets of transit skims to 3 for income classes
    • Get the overall integrated model setup working - DOS, EMME, Asim
  • Do all this and then run a 25% sample and re-share the visualizer next time

09/10/21

  • Discuss modeling needs with planning and policy staff
    • Will update model with latest travel survey results
    • And add TNC modes and work from home model (and telecommute frequency model)
    • Needs
      • Better model regionwide reduced fare scenarios
      • Need to output trip fare on trip records (like Oregon DOT also asked for)
      • Specify fare differences by income, income and HH size, and access/egress/transfer mode (especially new mobility modes)
      • Maybe we generalize the 3 to 5 transit assignment classes for general use?
      • How can we model a universal fare card across operators and fare groups?
      • Need highway VOT/income segmentation for pricing studies
      • Will be able to segment travel patterns for work from home versus not
      • Add Metra lot parking cost to MAZ-TAP input data and MAP-TAP expression file
      • Soon add race/ethnicity analysis; CMAP is working on adding it to the syn pop
  • Model setup
    • 100% full scale run now working
    • Needed to make some improvements to ActivitySim to support path building when not all path component data is available
    • Also improved the MAZ to TAP input data creation script
    • And trimming walk MAZ to TAP to 1 mi and drive MAZ to TAP to 5 mi
    • Cannot run on 512gb chunkless but runs by training 100,000 HHs and then running all HHs in production mode
    • Andrew updating mode choice - dropped toll/non-toll nest and added more transit access modes
    • Needs to add school bus
    • And then add three sets of transit skims by income
    • CMAP ok with running 50% samples and new servers will have 768gb RAM
    • I'll try a 50% run with 1 mi walk and 5mi drive max chunkless and see if it runs through

08/25/21

  • Visualizer now working with survey and activitysim model outputs for a 1000 HH sample
    • @Leah to make sure everything committed to the repo, including directions, for @CMAP to test
    • @Andrew will test as well before Leah heads out
    • Can we set a default drop down value @Binny/@Leah
  • Model setup
    • @Andrew add in work from home model
    • @Andrew working on updating the maz-maz and maz-tap script so at least one entry is available for each MAZ
    • Ran into issues with too few MAZ TAP pairs so no transit logsum and too many so ran out of memory
    • Working on fixing input data (and possibly ActivitySim) in order to get a 100% full sample run
    • We don't want ActivitySim to make bad choices...i.e. super low probability silly choices...so a max walk cap is still needed
    • Let's make sure school bus is a mode as well
  • Next steps
    • Get 100% full run working and update visualizer
    • Start working on issues/fixes to setup
    • Pause for model design/next steps discussion
    • Discuss policy needs with planners on 9/10 meeting

07/30/21

  • Asim setup
    • TVPB setup and working
    • Accessibility model setup and working
    • Ran into some memory issues but fixed them...basically needed to include the tapLines file to trim the MAZ to TAP pairs
    • Will add issue to ActivitySim to make this required (or at least warn about it)
    • Now working through rest of submodels for little translation issues
    • Some submodels running
    • Andrew out next week but plans to work on this the following week
    • Plan to run 100% sample with all skims on our 512GB machine or an Azure box with more RAM to inform CMAP's machine purchase
    • Will run with no chunking to see how much RAM we could ultimately use (in the base year)
    • There will be a new asim release in a couple weeks with improved chunking/memory that we'll want to use as well
  • Ct-ramp run with new inputs
    • We ran it with a small sample of HHs
    • We need to keep working on getting the full sample run through since this helps identify issues in the data
    • We'll switch to this data with the new asim setup
  • Updated summaries
    • Leah shared updated visualizer with final PACE boarding weighted targets
    • Will commit scripts and update readme
    • Will process asim outputs and add to the visualizer
    • Will check with Binny about a 3 zone system model outputs processing script, maybe for SANDAG or ODOT
  • Project schedule
    • Wrap up phase 1, this phase, end of August
    • Switch to Phase 2, revisions/fixes, for rest of year
    • Next year, calibration/estimation, etc
    • Invite policy folks to September 10 meeting

07/16/21

  • Asim setup
    • Andrew working on it
    • Has all the inputs converted and is working on getting 500 HHs running
    • Keeping track of issues/hacks as we go
    • Starting from SANDAG's in-development 3 zone system setup
    • Work is on the activitysim branch
  • New zone system
    • Some inconsistent projection issues across inputs that Nic will work on
    • CMAP share updated allstreets network
    • CMAP share updated work INDCEN file
  • Maz to maz / tap skimming
    • I have the script working and using pandana
    • Its very fast since using contraction hierarchies and multi-threading
    • I'm working on making it faster and finishing it for check-in
    • CMAP to share NAVTEQ network documentation
  • Leah processed the PACE data for the survey targets
    • Transit tour mode share went up a bit
    • She'll update the visualizer and share again
    • CMAP says the memo looks good
    • Next, get scripts working to process Asim outputs for the ABM visualizer
  • Project phases
    • 1 - get something up and running
    • 2 - refine/fix issues
    • 3 - calibrate/estimate/finalize
    • We're in phase 1 with the plan to switch to phase 2 later this summer once asim is running start to finish (with the list of issues/hacks still)
    • Craig bring policy folks back to meeting to discuss flexibility; probably mid August

06/18/21

  • Highway skims
    • Good enough for asim setup now
    • Scripting is being cleaned up and soon committed
  • Transit skims
    • Regenerate plots without NA paths to better see differences
    • Skims are tap to tap
    • Fares need work
    • Compare boardings by line
    • Compare transfer rate
    • Could use new MTC/Metro journey level fares code INRO wrote
      • or CMAP's similar code from the extended transit assignment python code they added
      • Requires operator fares and operator to operator transfer matrix
      • SF fares environment much more like Chicago than San Diego
      • Let's consider once we get the basic skimming up and running with asim
    • CMAP could also use the Metro/MTC scripts to code TAPs which are EMME/Python instead of SAS/ArcGIS
    • Andrew finishing this
  • HH survey / onboard survey
    • Updated mode targets with OBS and scaled with daily ridership
    • Walk transit went down a lot, walk went up
    • Maybe transportation advocate self selection bias and oversampling at rail stations?
    • Pace not included in OBS so need to add to weighting and rescale
    • Also create access/egress mode summaries from the OBS
    • Also cleaned up tour/trip mode combinations to match model design (i.e. exclude unlikely/funny combinations)
    • Not explicitly modeling bike transit, but could, but has pros and cons
    • Leah finish data processing, update visualizer, post scripts to repo next week
    • Craig make sure Leah and Andrew have invites
  • Setting up ActivitySim example
    • Will start with SEMCOG, which is a 1 zone system
    • Will pull in the SANDAG cross border travel model asim Network LOS 3 zone system and transit virtual path builder (TVPB) setup
    • Will merge these two to get the CMAP starting point
    • Will modify to work with CMAP syn pop, land use/zones, and draft TAZ and TAP skims
    • Ben and Leah working on this now
    • Leah on vacation for a few weeks starting at the end of next week
    • Send email update before July 2 cancelled meeting

06/04/21

  • CMAP model run on server
    • Asim example running on server
    • Will try with more processors and with MKL_THREADS setting
  • Updated zone system model setup
    • CMAP plans to share Monday/Tuesday and Andrew to setup and run at RSG
  • Survey processing
    • Updated visualizer shared
    • Day pattern shares looking a lot better
    • Reviewed other summaries and they look pretty good as well
    • VMT per HH increased 25%, which could be high, can we check against the model results?
    • Talked through the OBS data processing summaries
    • Metra survey had lots of missing data
    • Craig share 2016 Metra survey if helpful
    • Transfer rates by operator look ok
    • Metra has a lot of PNR
    • Now working on the improved tour mode shares using the scaling factors from the HH survey
    • Will update the visualizer for the improved tour and trip mode shares
    • Plan to wrap up next week
  • ActivitySim setup
    • Leah to then start setting up the activitysim setup with the syn pop, land use, and draft highway skims
    • Likely use the example_sandag 3 zone system setup
    • Leah out June 28 to July 14
  • Skims
    • See Andrew's slides, highway skim results looking pretty good
    • Need to check link congested times by facility type, especially expressways. Free flow times look ok.
    • Nagendra still working on the transit skims
    • Hope to get drafted next week and then Andrew can do summaries/QA/QC/fixes

05/21/21

  • Part two of ActivitySim training completed this morning
    • Would be good for CMAP to run ActivitySim on the server
  • Skimming
    • SANDAG EMME procedures now working in draft form with CMAP networks
    • Procedures are called from a Python script like the previous CMAP ABM setup
    • Reviewed some goodness-of-fit comparisons of the skims, assignment results
    • Need to work on VOT bin assignments and continue with the QA/QC work
    • A toll transponder model is unlikely to be added to ActivitySim soon since the technology is evolving
    • We can assume an EMME bank will be a scenario input so we can pull out just the create networks components of the macro that we need
  • Re-weighting
    • Re-weighting is helping, but not as much as desired since the distribution was already pretty extreme
    • Adding in kids under 5 helped
    • Also used the newer PUMS data for the re-weighting
    • Number of trips, stops, VMT all went up in new visualizer
    • Day patterns didn't move much though so Leah and Mark to discuss and fine tune if possible
    • Maybe add a tour factor
    • Can share populationsim-based weighting setup with CMAP
    • Need to wrap up re-weighting soon and move on to OBS mode shares analysis
  • CMAP todos
    • @Nick working on updated zone system model inputs
    • @Craig send network documentation as well

05/07/21

  • Part one of ActivitySim training completed this morning
    • I'll share the slides
    • CMAP staff to work on Getting Started guide and Getting Started notebook
    • Next training will be Q&A based on experience
  • Survey processing
    • We're making progress on the trip underreporting corrections
    • If no travel reported and retrieval by online then weights will be increased for trip makers and decreased for non-trip makers
    • No trips pattern for online goes down 50%, that's substantial
    • @Craig to look for NIRPC sampling plan to share with Leah
    • Will divide weights by average weight so t-stats easier to understand
    • Need to redo all the weighting as opposed to just additively weighting so we don't mess up other dimensions of the data
    • Visualizer not updated yet, but coming soon
    • Maybe add mode self selection correction factor too
  • For trip model
    • Using the syn pop and enumerating trip generation by sampling a trip generation rate for 624 potential groupings from the survey
    • Plan to use updated weighted averages for trip generation moving forward
  • Skimming
    • Have the SANDAG highway skimming toolbox working with CMAP networks, but only the basic skims - distance, time
    • @Craig to share network data dictionary if available
  • Updated zone system
    • @Nick working on it
    • ActivitySim MAZ data requirements very similar to previous model
    • Will send over updated model setup for us to run/use

04/23/21

  • Model setup
    • ran into a Java 7 versus Java 8 bug with unstable hashes in the existing model that wasn't in Java 7, but model now running at RSG
    • Working on the SANDAG highway and transit assignment/skimming toolbox migrations
  • ActivitySim training
    • trainings scheduled for Fridays 10 to 12
    • I'll send invites
  • Survey processing
    • Updated visualizer with correct time periods and two Lake counties
    • Replaced 3mi intrazonal distances with area based calculation
    • Drafted travel underreporting plan based on previous RSG work
    • About 1/3 of respondents used smartphones so that will be our base for the estimating accurate trip rates
    • Apply estimated underreporting model to factor up or down survey record weights
    • Do both day pattern corrections and nhb corrections
    • Deliver updated survey weights; should have for next meeting
    • May need to iterate a bit to get the kids trip gen rates right
    • start on OBS data processing after that
  • CMAP working on updated ABM setup with new zone system
    • We'll get a zone shapefile from the online data hub
  • Transit fare modeling
    • What if we gave low income a 50% pass discount? What's the impact on revenue?
    • RSG in the process of developing transit pass subsidy and transit pass ownership models for ActivitySim
    • Can be applied by market segment for different types of people/passes/operators/etc
    • Fare revenue estimation tends to be more of a network modeling question, like boardings
    • But with the person transit pass models, equity analysis becomes easier

04/09/21

  • Survey visualizer now up and running, including for both surveys
  • Tour rate is low
    • too many full time workers staying home
    • Needed to do a more convenience / targeted sample in the end to get enough sample
    • Let's calculate gps underreporting correction factor
    • Mark can work on this with Leah, but will need to be added to the contract
    • Only 60% of workers going to work, should be more like 70 to 80%
    • Results for students too low as well
  • Other survey processing revisions
    • Check intrazonal skim distances since something funny around 2 miles
    • Spot check some zones
    • Survey does not have some counties so update the worker expansion weights
    • Joint tours are low
    • Need to update results for the 8 time periods
    • Will get transit boarding numbers from the OBS for comparison
    • Will be good to compare demographics from the survey as well - zero worker HHs, HHs by size, age, etc.
  • Are in the process of calibrating the TBM with this data so there may be some boosting knowledge that we can reuse
    • Should be done within the next few months and will share results with this project team as well
  • Mark estimate trip generation model as a function of income, age, etc and then apply to identify trip underreporting bias
  • Next up, process on-board survey to calculate updated mode choice targets
  • Not much progress on the migration to the new EMME procedures since Andrew was on vacation
  • Asim training after next asim release
    • I'll create a doodle poll for scheduling
    • 2 hour training, then homework, then another 2 hour training

03/26/21

  • CMAP Model on RSG Server

    • Andrew and Leah downloaded the existing CMAP model on RSG's server
    • Andrew configured and started a run
    • Will check the outputs next week

  • Network Skimming Procedures

    • Andrew just getting started on updating network skimming procedures

  • HTS Processing

    • Leah ran the CMAP HTS through the SPA (survey processing application) tool
    • Leah/Binny described the SPA algorithm, person type coding logic, and mode and purpose crosswalks
    • Initial summaries show a weighted person tour rate of 1.28 persons per tour and a household trip rate of 8 trips per household
      • The tour and trip rates are slightly on the lower side. Typically a person tour rate of 1.35 and a household trip rate of 9-10 is expected. This could be due to the underestimation of stops. The survey also doesn't report trips for all preschoolers. Craig noted that some respondents reported 'not traveling' on the survey day which may not be true.
      • The survey respondents had the option to use GPS-app, or complete the survey online or via phone. The survey mode variable is not available
    • The mode and purpose summaries by trip and tours seem reasonable. The transit mode share is ~6%. Leah will check on Loop trips/tours. TNC trip share is ~0.7%
    • The trip mode X tour mode summary may include combinations not allowed in the model (e.g., SOV trip on Transit tours). The mode combinations that are not allowed will be reallocated when creating mode choice targets for calibration. The mode choice targets will also incorporate transit information from the on-board survey. Craig will share the OBS if not shared already
    • The person type distribution looks reasonable. Leah will update SPA to include preschoolers. Part-time worker share seems a bit high. Leah will try a 30-hour per week threshold to code part-time status and compare it against the 35-hour threshold
    • Will stand up the ABM Visualizer next and perform QA/QC on HTS processing

  • CMAP Update

    • Posted PopulationSim setup to GitHub
    • Preparing input files for setup with the new zone system
      • Done with synthetic population
      • subzone and landuse file in progress
      • Will get started on Emmebank after next week

03/12/21

  • Leah started on the survey processing
    • Will join the surveys and use the ipf weights
    • Starting with the MWCOG ActivitySim Python SPA scripts
    • Plan to get the data in order and setup the survey visualizer for the team to review
  • Andrew just starting on the network modeling updates
  • I need to work on finalizing the plan
  • Craig shared model setup on CMAP server
    • Leah to download and run on our server
  • Craig shared a ranking of policy needs for motorist fee projects:
    • VMT charge based on income
    • Vehicle type charge (gas versus EV) - maybe a probability lookup table based on household and trip characteristics and based on the travel survey
    • Cordon pricing (charge for operating within area) - would be good to get more details so we can figure out how to code scenarios in the model
    • Carbon tax
    • Cordon pricing (charge to enter)
    • Time-of-day charge

03/05/21

  • @CMAP working on getting new zone system data together
  • @CMAP is going to grant us access to their server for running the model
  • Discussed model update plan (restricted access)
  • Will TNC to transit be in the model? Yes it can, although it can get complicated and so we'll need to sort out the design once we get the initial system up and running
  • Plan looks good so we can get started
  • @Ben to finalize plan
  • @Leah to get started on the survey data processing
  • @Craig to share the updated survey files with some additional data cleaning
  • @Andrew to get started on the EMME network modeling update procedures
  • @Binny to provide support

02/26/21

  • We'll use this GitHub site for managing and running the project
  • Model inputs
    • Craig send latest version of the ABM to RSG, as well as the HH survey, and new zone system
    • Will need to sign NDA
    • CMAP to update inputs to use new zone system - land use, syn pop, networks
  • Discussed work-in-progress model update plan (restricted access)
    • I'm putting the revisions into the doc, but here are some key points:
    • keep TAPs and time periods
    • Assign highway user classes by VOT bins but also create VOT+income bin matrices for post-processing
    • Assign transit user classes by income categories
    • Should we swap out the existing macro-based highway and transit procedures with SANDAG's new EMME modeller toolboxes?
    • Should we do multiple transit path sets or just one (all modes) like before? Can we test various transit path building options to select the best approach?
    • We plan to keep the transit capacity, crowding, and reliability approach, but maybe try new the EMME stuff (like MTC)
    • How should we phase the model transition? Asim setup first, network improvements, etc?
    • Need to talk through the accessibilities approach next week
  • Let's talk again next week so we can keep the project moving along

02/12/21

  • Project kickoff meeting
  • Key model needs / requirements
    • Equity and inclusive growth (often defined using Census GIS SES data)
    • Transit subsidies by operator, TOD, users (income)
    • Road user charges by income
    • Use ABM for large policies
  • SEMCOG Asim model recently built
    • Started with TAZ level model and now switching to 30 minutes, 3 zones, adding telecommute, TNCs
  • Income class assignments good for policy analysis
    • Both highway and transit
    • Could do VOT classes instead and back into (post-process) income shares
  • Likely keep 8 time periods for highway and transit
  • TAPs
    • Drop all the station coding stuff except station type (pole, shelter, station, etc...)
    • Dropping TAPs means a smaller EMME license
    • But TAPs good for transit modeling so maybe keep?
    • Could try the DaySim 2 zone approach (MAZs and TAZs only) with some improvements to better model walk access time
    • TAPs important for transit project analysis
    • Could maybe simulate transit walk times like before?
  • 5500 taps, 3600 tazs
  • Keep MAZs and active transportation / allstreets
  • Accessibilities and nearby MAZ impedances
    • Re-implement custom CMAP code to asim calculator
    • Re-implement custom CAMP code to python - WFRC micromobility python code or PSRC pandanas or EMME
    • Check on EMME enhanced network model cost for this feature
  • Asim setup
    • Start with asim example and switch out land use, syn pop, and modes
    • Then go through each submodel and build a comparison table
  • Toll assignments
    • Not enough toll users in assignment results
    • Drop toll/non-toll in mode choice and switch to VOT classes
    • We switched to SOLA assignment as well
    • Path based assignment is good for select link analysis
  • Macros
    • Update to Python as it makes sense but not a high priority
  • Set up an Asim training, 2 to 4 hours
  • Make sure move to new zone system is in the update
  • CMAP already using populationsim
  • We will use the Asim SPA Python survey processing code and RSG survey visualizer
  • We can look into Replica and IL statewide model external flows, district flows later during validation
  • Let's start on the EMME network modeling updates earlier rather than later
  • Craig talk with policy folks again
  • I'll work on a model update plan
  • We'll schedule a bi-weekly meeting
Add a custom footer
Add a custom sidebar

Clone this wiki locally