Skip to content

Project Meeting 2018.06.21

joecastiglione edited this page Jun 25, 2018 · 26 revisions

Stop models

  • Trip mode choice implemented in branch trip-mode
  • Updated trip destination choice to use trip mode choice logsums now as well
  • Needed to revise and clean-up mode choice utility specifications to be easier to use / finalize
  • @Jeff working on the full scale example run now
    • A few trips failing with all zero_probs, and still investigating
    • Issue in tour_mode_choice caused by the fact that some WALKDIST and BIKEDIST skims are asymmetrical (O->D vs. D->O) so tour_mode_choice would allow outbound walk (<3mi) but the return was >3mi, so they couldn’t get back
    • Remember transit subzones were not implemented in this version
  • Once complete, we'll do a pull request and then update documentation and re-release

Documentation and release

  • Example model design and sub-model designs updated
  • Sub-models hyperlinked flow chart integrated
  • Updated listing of all sub-model input files
  • Updated data schema and built auto-documentation tool based on our data pipelining technology
  • Updated getting started steps
  • Lots of little clean-up in the docstrings and existing documentation
  • Released Friday June 8th
  • @Ben will update and re-release for trip mode choice updates

Project management

  • Updated Progress Report through trip mode choice
  • Current scope will be completed in a few days and all funds will be expended by then
  • Updated Phase 4 Scope of Work posted for comment
    • Moved multiprocessing to the top since it is the next important step
    • Updated shadow pricing design to be flexible for use in all regions - allow for specifying classifiers for combinations of persons, land use, and geographies in the form of expression files for example
    • Moved continued verification and clean-up to the end since it's an ongoing part of the overall effort
    • Moved multiple zone systems to the end and suggested we test with either MTC TM2 or SANDAG instead of TM1 + fake data
  • Discussed how best to implement support for multiple zones systems since it is not required for every partner
    • No one wants to create fake data for the TM1 example
    • @Lisa and @Wu to think about setting up MTC TM2 and SANDAG as the test example for this feature
    • @Ben to break out task into sub-tasks
      • Setup example/test to implement new features for, including creating all required inputs
      • Implement support for multiple zone systems, including updating expressions in all relevant sub-models
      • Implement transit virtual path building, including maybe a simple version first and then the more advanced version for TM2 and SANDAG
    • See table below for zone systems in use by region
    • Should we do something else in the Phase 4 SOW instead of multiple zone systems?
    • It might helpful to review the multiple zone system design
    • We want to make sure to keep, and test against, both examples - TM1 and the multiple zone system model example
  • See you Monday night in Atlanta

Zone Systems in Use

Region TAZs (1) TAZs + MAZs/parcels (2) TAZs + MAZs/parcels + transit stops (3) Notes
PSRC X Transit IVT, etc from TAZ-to-TAZ skims. Transit access/egress times looked up by sub-mode in the buffered MAZ/parcel file and represent time to nearest stop.
SFCTA X Transit IVT, etc from TAZ-to-TAZ skims. Transit access/egress times looked up by sub-mode in the buffered MAZ/parcel file and represent time to nearest stop.
MTC TM1 X
MTC TM2 X Does transit virtual path building during a model run to build MAZ to MAZ impedance through the best transit access and egress stops
SANDAG X Does transit virtual path building during a model run to build MAZ to MAZ impedance through the best transit access and egress stops
SEMCOG ? ? ? Trip-based model uses TAZs
ARC X

Comments

  • ( Joe ) It would be helpful to have estimates of level of effort (hours/dollars) by task (and possibly subtask, for multiple zone system task) as we try to refine scopes and prioritize.
Clone this wiki locally