Project Meeting 2022.04.21

Agenda

Code for the vehicle type model is all implemented and they are wrapping up documentation. RSG will report to the group later for a target completion/delivery date.
Testing will include running for MTC and making sure that the vehicle distribution is close to NHTS.
RSG is currently investigating sources to come up with auto operating costs, which will vary by vehicle type, year, and power type.
- AAA data used to generate annual report, that data is proprietary and not able to find other useful sources.
- Going through published AAA Know Your Driving Cost reports and compiling that information into a table to populate for the model.
- Question about historical data availability, can you get auto operating costs for vehicles up to 20 years old? Not sure, need to ask David
- Some issue with that data - based on analysis of top 5 vehicles sold that year, not representative of the entire fleet
Sensitivity testing
- Greg had done a lot of sensitivity testing.
- Additionally, RSG will propose a couple of sensitivity tests for the group.
Question about how to maintain consistency between upstream model components and assignment is not in the scope of this project, but might be a good topic for later.

The visualization branch is merged with the main ActivitySim repository, effectively closing out this task.

Code changes related to benchmarking have not been merged yet. There is one parameter that fails the Travis test; Jeff is working to resolve this outstanding issue.

Jeff is testing a poll for prioritizing data dictionary issue: https://github.com/ActivitySim/activitysim/discussions/551
- Propose to standardize data dictionary/metadata files which would include the following benefits:
  - Help with visualization to have all the metadata available and not have to manually specify
  - Would be included in Github and not have to look into wiki, ensure version control
  - Potentially low technical effort but high value to the users
Another example application of the polling capabilities of Github discussions would be resolution of the bit issue. During a meeting, this issue wasn't resolved but the group was asked to resolve via Github issues, but no one added anything. Maybe polled discussions could be a better way to handle such issues?