Skip to content

Project Meeting 2022.08.11

mnbina edited this page Aug 11, 2022 · 6 revisions

Admin

Web content due today

  • Received already from CMAP, SEMCOG, Oregon DOT
  • SANDAG sent in content and will double check with Caitlin that it was received.

Payment for current year also due soon

  • Payment received for Oregon DOT, MWCOG, and CMAP. If you have submitted a hard copy check and are not listed above that may be because they have not yet received it or it has not been processed by our accountant. Once received and processed, you should receive an email from they accounting assistant (Deb) confirming receipt and Caitlin will continue to update the group over the coming weeks.

Release of Version 1.1 is almost ready.

  • Renaming of examples is done
  • Comprehensive full scale testing is nearly done
    • All models in the examples manifest are being run
    • Tests exclude vehicle type model, that’s in the extended model.
    • The full extended model is the kitchen sink example that includes all the new features since the version 1 model. Not feasible to add every new feature to every example model. MTC Example is frozen, as the stable initial version. Provides a guaranteed benchmark to compare all new updates/versions. All enhancements go to the extended version.

Discussion

Differences in escort model between proposed ActivitySim solution and Tourcast.

  • Tourcast
    • The choice of escorting is made simultaneously for outbound and inbound legs of each child, and sequentially over children from youngest to oldest.
    • Older children are bundled with younger children by availability constraints on alternatives (e.g. if older child school tour leg is same time same place as a previously assigned younger child, they are bundled, otherwise not)
    • Availability of alternatives is based on previously determined schedule of mandatory tours – max 90 minute time adjustment allowed
  • Proposed ActivitySim:
    • The choice of escorting is made simultaneously for all children (up to 3), and sequentially across legs (outbound, inbound, reconsider outbound).
    • Availability of alternatives is based on previously determined schedule of mandatory tours
  • Main difference is: Tourcast is sequential and ActivitySim is simultaneous in terms of bundling alternatives.
  • Essentially same alternatives, question about any differences in downstream models and how it’s handled
  • ACTION ITEM: CS to check to see if all school escorting is assuming to be auto in TourCast?

Do downstream models impacted by the school escorting implementation need to be recalibrated?

  • Original scope anticipated that there would be downstream impacts.
    • What is the level of effort to recalibrate, how much have things changed?
    • As a general principle, as we build new components, we want to evaluate whether or not changes produce crazy values or not. We want someone to be able to pull something off the shelf and that it’ll produce reasonable results.
  • RSG previously implemented for SANDAG, and they adjusted non-mandatory frequency, stop frequency models, and then redid probability distributions for stop purpose that are effected. They could transfer directly from SANDAG. They don’t have MTC regional survey data with these tours/trips coded, so they can’t do calibrated from local data for the MTC example.
    • To be completely consistent, you’d want to re-estimate these models. Downstream models have interaction terms such as children in the household, which may not be relevant anymore if those tours aren’t included in the non-mandatory frequency model, for example. These tours were included in the estimation, so they may not be significant anymore if they were excluded from the estimation dataset. RSG will be re-estimating these models for SANDAG, but the timing for that is not consistent with this task order.
  • When scope was originally proposed, work was being done for SANDAG and would be transferred to MTC. What’s changed is that there is an increased sensitivity that there should be more than light recalibration, re-estimation is recommended.
  • Want to wrap up work applying the existing scope. Let’s pull what we can pull in and then assess what the degree of recalibration is needed – it could be zero or more than zero. Hard to guess what it would look like without seeing how far off things are.
    • Compare of number of tours and trips, for example
    • Only assessing with the MTC example model
  • ACTION ITEM: Add to next week's agenda:
    • What is the reason for introducing this model at this time to ActivitySim. Can we wait to roll out a version that includes this model until we can re-estimate? Will this be part of a new version?
Clone this wiki locally