Project Meeting 2022.03.17

Action Items

• Members to discuss options for AMPO Conference presentation, which is due April 1
• Check back in with Sijia and WSP on how the testing

Admin

• Phase 6 updates
  • Benchmarking:
    • Jeff addressed RSG review comments, waiting for RSG to approve responses
  • Visualization pull requests, passing request, waiting on code review on that
    • Elizabeth to code review
    • Joe Flood to code review
  • Vehicle type work still in progress
• Issues tracker
  • ARC example model –issues in trip distribution related to skims, placeholder fix earlier to get the prototype running, needs a more comprehensive will be fixed later
• AMPO conference
  • Need to submit abstract – due April 1
  • Desire to propose something more than a just a typical presentation, since there is so much to cover
  • You can ask for how much time you need, maybe 60 minutes, possibly 120 (unsure about the options available), you can add a panel or group of presenters
  • Caitlin said to specifically address what the AMPO conference attendees will take away from the presentation and recommend interactive components (polling, small group discussion, multiple speakers, interspersed Q&A, etc.)
  • To be further discussed at the members only meeting later today

Broader Testing (presentation by Sijia @ WSP)

• Scripts to run individual model components
  • You do need to have run upstream components first (or have data that includes the upstream information). This could come from a different run (like from the java-based outputs).
  • Similar issue during benchmarking activity, here was Jeff’s process to account for this issue:
    • Run full model to use ActivitySim to recreate the check points, generate all the data which takes time, but then you have all that data and is necessary to generate for this kind of testing.
    • Need to enable checkpoints to have it to write to disk. There currently isn't code to write only the marginal data.
    • Recommends writing out what each step needs as an input and what is being output in order to remove anything extra along the way
    • Suggestion of having a data validation pre-process to check the inputs and make sure it’s all in the right format, etc.
    • This could be a significant effort to write the code and better model documentation.
    • What's proposed is to streamline the pipeline but may lead to some redesign.
  • Next steps
    • Put on the table as something that would will be necessary but not capable to do right now, likely would have to wait until Phase 8.
    • WSP will continue to develop their testing pys for each model component.
    • It was proposed that some T&M flex partner resources be used to spec out and design this.