Project Meeting 2022.10.20

Agenda

• Code Review/PR Assignment
• Update on shadow pricing task (RSG)

Action Items

• All Pull Requests were assigned to reviewers in Github.
• Michelle to follow-up with Jeff on making the AMPO presentations available to the group.
• Joe to look into Daysim approach for handling out-commuting workers.

Meeting Notes

Code Review / Pull Request Assignment

• School Escorting – Joe Flood is reviewing the outputs, not has not looked at the code itself. He volunteered to also look through the code. Estimated level of effort: multiple days.
• Flexible Number of Tour and Trip IDs – Joe Flood to review. Estimated level of effort: half a day.
• SEMCOG Example - Jeff Newman to review. Changes are primarily configs, but the primary review should be to run the tests and full model. There is no need to go through the the code or configs at this time. Estimated level of effort: one day.
  • Folks questioned about whether new/updated examples should be reviewed. One one hand, if there is no code pulled into the ActivitySim code, nothing should affect other users and there’d be no need to code review. On the other hand, not fully testing an example model did cause some problems for you when you were testing sharrow, so there is some value in making sure the examples are working correctly for testing new code across several examples.
  • Question about whether Jeff moved the SEMCOG extensions into the ActivitySim code, which will be confirmed. If they were moved into the ActivitySim code, it would likely need to be reviewed.
• Disaggregate Accessibilities - Sijia to review. Estimated level of effort: a couple of days.
• Shadow Pricing - tentatively Sijia to review. Estimated level of effort: a couple of days.
• Skim Wrapper - Jeff Newman to review. Estimated level of effort: minutes.
• Sharrow- Sijia and David to review. Estimated level of effort: low.
• PTV's Window Installer – Jeff to review.
• Estimation Fix - David to review.
• Numpy Versioning – these will need to changed soon anyway, so holding off on this.
• Random Seed Generator - David to review. Looks like it’s suggested to do a little more development to create a test, to be done before review.

Shadow Pricing Updates

• Presentation: shadow_price_comparisons_20oct22.pptx
• Convergence
• Added Daysim tests for performance comparison, very similar to CT-RAMP
• Tweaked convergence criteria
• Concern about issue in slide of workplace simulation convergence chart, results look odd and needs further investigation.
• Removing WFH/External Workers
• Lots of ways to address the issues and an approach was chosen that allows flexibility for user to specify their preferred approach. This could even be applied on the school side, if there was a university with significant in-commuting, for example.
• Added factor in landuse file that specifies WFH (which is actually in-commuting workers) shares and then scaled employment (where user could make it more sophisticated by specifying different scaling factors by industry). RSG will replace WFH reference here to in-commuting workers.
• There was a question about workers with the external jobs.
  • SANDAG model includes a component that specifies the number of workers that have an external workplace, they are removed from the choice set and another routine is run that assigns them to an external station. RSG has presented on this before.
  • Is there a simpler approach that can be applied here? Can we do the inverse of this approach for out-commuting workers. This needs to be thought through for downstream effects. What would you do for the out-commuters in daily activity patterns, assume no travel? Joe to look at processes in DaySim to look to see if there is a simple solution.
• Targets
• Current targets are set for total enrollment, not segmented by grade. This could lead to university students switching to a grade school that was underestimated.
• Note that there is no grade school enrollment info in the MTC example.
• Agreement that segmentation is the way to go but that the MTC example isn’t the best way to test this.