Project Meeting 2021.02.09 - ActivitySim/activitysim GitHub Wiki

Technical Call

  • Recent issues review and PRs
    • Re: #377, including Newman's idea to temporarily disable trip-level processing for tours based on tour mode
    • Lots of discussion around Newman's idea, which we can discuss in more depth later. We want to make sure to implement features like this in a way that are transparent, optional, and do good modeling.
    • Stefan suggested add reading skim data from disk on-demand as opposed to reading every skim into RAM at the start as a way to trade runtime for RAM. Jeff implemented an undocumented version of this during the TVPB caching research and it runs slower but uses a lot less RAM. We may want to complete this feature for general use.
    • Re: #376, the number of work tours could be configurable but it may be difficult to maintain/manage the code as a result. You need to be careful coding work trips for commercial work trips (like plumbers, etc. that travel each day as part of work)
    • Re: PR #372, we may want to improve the output trip writer to support annotations as a way to support more flexible/custom output writing
    • Re: #373, Instead of making these rigorously consistent by being defined together, we could just throw a warning message if they are not consistent. DTA integration is an example where we might not want the skim time periods to be the same as the trip matrix time periods.
  • Discuss performance task, taz/maz location sampling progress for PSRC
    • Mostly working now and Jeff creating a before/after runtimes comparison for review
    • The iterative trip purpose and destination model was causing a few failed trips
    • Maybe we want to create an override choice option after max iter reached? I think we already do this, see the config file
    • Jeff to finish up and then commit and Stefan to test
    • Jeff has ARC example running and will turn to representative logsums next
  • Discuss ARC extensions additions and improvements
    • Joel to share list of improvements/fixes from SEMCOG application of the ARC half hour tour TOD choice model
    • I asked for an example of the ALTS_PREPROCESSOR used in the ARC mandatory_tour_scheduling model for the test system so we have a documented example
    • However, we think upcoming Task 6 will help us solve this problem so Clint to wait on contributions until we get into that task
    • ARC also thinking about how to write trip matrices...either with Cube scripts that read trip files or with ActivitySim's write trip matrices function
  • Discuss next time
    • Discuss estimation enhancements
    • SANDAG accessibility testing memory issues