Project Meeting 2022.07.19 - ActivitySim/activitysim GitHub Wiki

Agenda

Meeting Notes

  • TAP Approach
    • Motivations
      • If TAZs are large, distorts access to transit stops
      • Opportunity to introduce alternative specific constants
    • How it works
      • Generating access and egress points between MAZs and TAPs with Python software.
      • Transportation software to generate transit skims between TAPs.
      • Custom software to combine.
    • Pros and cons of TAPs (in presentation)
  • Alternative to TAPs – using TAZs and MAZs only
    • How it works
      • TAZ assignment
      • Using MAZ information to augment it
      • Comparison to TAP approach by utilizing MAZ information
    • Pros and cons of 2 zones (in presentation)
  • Test to see if the two approaches produce similar results
    • Plotted out TAP and non-TAP travel times for O-D transit trips from onboard survey data
    • Not strong bias one way or another
    • Non-TAP approach results in high auxiliary walk time and slightly lower times for other components
    • With more effort, adjustments could be made to make results closer together. Ideally, both approaches would have more calibration.
  • Another test
    • Using example TAZ near Jack London Square
      • Using TAZs only, about 5 minute impedance for walk access
      • For a far away MAZ, the walk time is 12.5 minutes to the BART station, which is longer than the 5 minutes in the non-TAP approach
    • How often is this happening
      • If there is only one transit stop, then the non-TAP approach will be same as the TAP approach
      • If there are two stops that are equidistant to the MAZs, then the non-TAP approach will be same as the TAP approach
      • If there are at least two TAPs and the distance differences are greater than 0.25 and there is some transit demand, the difference in approaches impacts about 35% of all MAZs.
      • Considering only MAZs with significant transit demand (more than 100 walk to transit trips) then you are down to 1%. Even less common were the instances where the demand is significantly different (0%).
  • Spatial distortions caused by TAPs
    • If the TAPs include stops far apart, the TAP approach could show a much shorter access than actually exists.
    • Similarly asking how often this happens, 2.8% of MAZs that have connections that are more than 1 mile away. Even less common for those with significant transit demand (0.4%).
  • Conclusions
    • A non-TAP approach is viable and produces similar results to TAP approach
    • TAP approach has some benefits
  • Recommendations
    • Don’t use TAPs if your region has many more transit stops than TAZs
    • Benefits of TAPs are compromised when more than one transit stop is connected to a single TAP.
    • TAPs make more sense for those agencies that are more comfortable with the concepts, technical mechanics, software, etc.
  • Another idea
    • Develop Transit TAZs which are aggregations of MAZs based on transit access
    • Roadway and transit TAZs wouldn’t be the same
    • Use lookup/indexing, would require additional custom coding but may not be as intensive as the TAP system
  • Comments from group
    • SANDAG will move to 2 zone system, based on this analysis and their internal discussions. Reason behind their decision is because it is so computational intensive. The custom coding is hard to maintain, and they want to get rid of it. (They do however have fewer transit stops than TAZs.)
    • SANDAG also thinks that the decision should be made considering network coding. They are currently going to switch network coding so now is a good time for them to switch.
    • Concerns about stochasticity in the 2-zone approach that uses assignment information. There may be ways to minimize it but would that approach increase run-time; need to consider the tradeoffs.
    • This comparison was focused on comparing the approaches but does not consider the implications for implementing in ActivitySim.
    • 2 zone is easier for social equity analysis when it comes to transit path building (consider transit fares)