Project Meeting 2020.12.15 - ActivitySim/activitysim GitHub Wiki

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