Project Meeting 2020.11.12

5b/6a Task 1 Project Management (milestone)

Oregon DOT is planning to multiprocess populationsim and add seed sharing for their statewide model and contribute the improvement
- Stefan reminded the team to say multiprocessing instead of multithreading and shared a link
- The statewide model runs multiple years into the future and runs for 50 to 60 PUMAs
- Plan to multiprocess by PUMA
- MetCouncil used seed sharing to add additional nearby PUMS samples to rural areas with not enough records in small markets
- How do we decide how to share seeds - use neighboring PUMAs?
- What's the process for rogue contributions like this? How do they align with our roadmap and the managing contributions section of the Strategic Plan?
- These contributions sound good and are from trusted sources
- Folks interested in the project should reach out to Alex
- Let's use this as a test case for a more formal review process
Ohio DOT is in the process of joining!
As the user community grows, it might be good to do a semi-annual update for the broader community
And it would be good to align that with versions

For the full scale Marin TM2 example, I corrected the drive transit issue and have traced a couple ODs to check results
We added ability to re-trace tap-tap calcs when pre-computed since otherwise they do not get traced
I mapped the selected TAPs for the trace and they look good; also the trace calculations look good
I'm working through the set of review summaries we discussed
TM2 TVPB coefficients were estimated by assigning the onboard survey and comparing paths
Can we add a total utility column to the tracing? Maybe we add multiple levels of tracing for general and then detailed analysis. Should all time periods be in the tap-to-tap trace or just the relevant one?
Maybe we exponentiate ahead of time to save runtime
Jeff added a new submodel step - initialize_los - that pre-computes the tap-to-tap data
He's working on the shared memory data structures/access revisions for multiprocessing now
Inside mode choice we'll just be ranking best paths - not calculating all the detailed path utilities again and again
We can discuss the DaySim PNR lot choice with ActivitySim page I drafted next time

Newman drafted the non-mandatory tour frequency interaction_simulate larch notebook
Newman working on wrapping it with a run_all_steps feature and also revising how the 8 EDBs by person type should be merged into 1 EDB like the other submodels

Telecommute design and links to background docs posted
Generally looks good
Can we make it any easier for the user to do what-if analysis? Maybe some kind of smart constant calculator for target telecommute shares or something like that?
I plan to draft spec/example for Blake to start programming
Any additional comments due next week

Joe talked with Beam/LBL, BYU, and hasn't heard back yet from Toledo
Caliper is helping Toledo with their setup
Joe plans to set up a call with these users to learn more about their experience
Need to review Clint's ARC CDAP and Duplicate Time Labels PR
Need to review Albab's SEMCOG joint_tour_destination configs PR