Project Meeting 2020.12.15

Technical Meeting

• Talk with BYU Professor Greg Macfarlane about his experience with ActivitySim
  • Using asim for an UDOT research project on potential TNC partnership with UDOT for wheelchair accessible travel
  • How many TNC cars needed?
  • Using popsim and matsim too
  • Using skims from WFRC regional model
  • Added wheelchair/disability status to synthetic population from PUMS and NHTS data
  • Would appreciate more stable versioning / releases
  • Would appreciate more documentation and access to helper resources / scripts / etc.
  • Would appreciate better release notes
  • Dealing with the skims was an issue, especially non-consecutive zone ids
  • Model is 2.5 mi people and runs in a few hours
  • Need better documentation on chunking
  • Would appreciate more advice on setting up for your problem - population size, computing resources, etc.
  • Increase uniformity - make a system that can more easily be transferred
  • Create other template scenarios for more typical regions than San Francisco
  • For example a Sioux falls model like for traffic assignment
  • A region more like most MPOs - i.e. just one bus service mode (i.e. no premium/rail service)
  • 15 MPOs have rail, 300 don't?
  • Most ABM users are big agencies currently
  • Cities like Roanoke, Asheville, Logan could all use this model
  • Need to focus more on documentation - inputs, configs, tutorials, etc.
  • Asim is wonderful compared to previous ABM software
  • Software implementation is really good
  • Add a popsim national/generic/Census example that works out-of-the-box for region
  • Doppelganger has this
  • Oregon could be a good place to develop the simpler example
  • Now working on beam setup (beam is SF improved version of matsim)
  • asim does initial plans and beam replans
• Update from Clint on ARC
  • Rebased ARC to multizone branch to get latest develop code
  • Will PR trip scheduling choice, trip departure choice, and trip parking location choice submodels
  • Now running full ARC example
  • Hanging on reading skims and so Clint to talk with Doyle
  • The memory footprint is now a lot smaller
  • Should be using new dynamic chunking
  • Jeff explained rowsize - how many columns for the row
  • Jeff wants to do more dynamic chunking improvements under the performance tuning task next time
• Update from Jeff on adding estimation mode to the trip models
  • Stop frequency submodel now updated
  • Working on trip destination now
  • Trip purpose is a lookup table of probabilities and so there's no estimation
  • Trip scheduling is a lookup table of probabilities and so there's no estimation
  • Will work on trip mode choice next week
  • Moved to new branch - estimation
  • Newman work on cdap and nmtf updates next week
• Update from me on the Marin TM2 work tour mode choice verification
  • Just about have all the expressions coded and debugged
  • Something not right still with drive transit that I'm investigating
• Update from Stefan on PSRC testing
  • Ran into a memory issue trying parcel to parcel aggregate accessibilities and so switching to maz to maz
  • The TM1 aggregate accessibilities calculator was TAZ level and we don't have a TM2 40k mazs test example, but Marin works ok
  • DaySim doesn't do aggregate accessibility calculations since users use a separate tool or script
  • Stefan to work on this a bit more and talk with us as needed