Project Meeting 2021.12.07 - ActivitySim/activitysim GitHub Wiki

Agenda

  • Update on chunking tests with ARC

    • Have not seen the improvement hoped for and more investigation is needed.
    • There are a lot of run settings that can be adjusted, e.g. chunking, turning off checkpointing, etc.
    • Propose a more systematic evaluation leading to more exact documentation and recommendations
  • Next Steps on vehicle type and allocation model

    • Utility specification for vehicle type model now acts more like CDAP or location choice
      • Non-mandatory tour frequency was the model framework
    • Model estimation & implementation
      • Propose to assess goodness of fit for model 4. Can compare NHTS distributions to model output. Want to ensure that the model is implemented correctly and there are no coefficients egregiously missing.
      • Implementing in a region where an agency that has interest or active development in this work would be preferred. Using MTC as the "canonical example" also makes sense. Would like a decision by end of next week.
      • For vehicle "allocation" (deciding which vehicle is used in each tour), will need to code ActivitySim-like tours in NHTS and estimate the chosen vehicle for the tour. Believe this would perform better than a more simple heuristic approach. Would not expect this to vary much between regions.
      • Would have to think about how to handle vehicle choice in the tour/trip mode choice file.
  • RFP review and questions

    • Request to move to a 12 month development cycle vs 6 month cycle
      • Flexibility for timeline and budget is already built into the RFP.
      • Want a strong connection between funds put into project and completed work is important. To that end, having a 7b in by June 30th for the end of the fiscal year is important.
  • Please send in your visualizer requests!