Project Meeting 2020.04.03 - ActivitySim/activitysim GitHub Wiki

Task 1 Project Management (milestone)

  • Discussed draft Phase 6 scoping Google doc here
  • Comments due in two weeks, 4/17
  • Plan to begin to prioritize in second half of April
  • Make sure we get this process into the strategic development plan

Task 2 Strategic Planning (milestone)

  • Discuss strategic development and contribution plan outline here
  • Comments due in two weeks, 4/17

Task 4 I/O Improvements (milestone)

  • No updates

Task 5 TNCs (milestone)

  • Summarized updated TNC example from MTC
  • Updated ActivitySim expressions
  • Need to run full sample and compare
  • Use the verification summaries from last time

Task 6 Model Developer Tutorial (milestone)

  • No updates

Task 9 Estimation Mode (milestone)

  • Discussed progress to date:
    • New estimation.yaml config file and estimation example being built
    • Each sub model gets an override choice inputs table with either HHs, persons, tour, or trips and a choice column override
    • Lots of work done to clean-up expressions so coefficients are separate from data
    • Working with sanitized data so we don't use resources cleaning specific survey data
    • Software is quite agnostic with respect to data fields so functionality is flexible
    • User will input a household, person, tour, trip table and activitysim can also output the same thing (it currently outputs this + additional intermediate data fields as well)
    • Estimation data bundle (EDB) built for auto ownership, work location, tour mode choice
    • EDB includes coefficient file, model specification file, data values combined file (chooser data and alternatives data), and choices file
    • Working through all submodel now and restructuring inputs + adding estimation integration code
    • Python estimation tools reviewed and going forward with larch
  • Next steps
    • Jeff continue to work though code, Ben to work on estimation integration scripts
    • About 2/3 of task budget spent
    • How much bad data, error checking will be included? Do we toss bad records? For now plan is to have more articulate logging around input data issues.