Project Meeting 2020.12.03 - ActivitySim/activitysim GitHub Wiki

5b/6a Task 1 Project Management (milestone)

ActivitySim.org domain renewal is up and we're going to ask AMPO to take this over. @Ben to forward email.
New members
- MWCOG and Ohio DOT both invited to meetings
- Both expressed some concern with the amount of meeting time
- MWCOG may wait to join until they have a model up and running
- There's interest in ensuring agency users join the consortium to help ensure a high quality and consistent product for everyone involved
- They also get the opportunity to participate in the project leadership
- Meetings are totally optional and many folks only participate occasionally
SANDAG cross border model
- RSG has now started on the model coding, which will use the 3 zone code and TVPB
- We'll get it working and let the consortium know how it goes
- We may end up making some improvements / pull requests as well, let's see
- Project needs to be done by end of 2021 but we plan for a running example early 2021
- Plan to use estimation mode as well

5b Task 2 Strategic Planning (milestone)

Joe setup doodle poll for agency partners to discuss latest revisions
Plan to get comments back to RSG later next week

5b Task 10 Support for Three Zone Systems and TVPB (milestone)

Multiprocessing working well
Now working on updating the documentation, example, verification
Below are some max maz-tap walk distance threshold tests with expected results that we discussed
We'll include a discussion of results like this in the documentation since it's important for understanding how to setup the model and how to code the network
Can use the household travel survey to determine a good max walk cutoff

###### tap to maz distance cutoff sensitivity
 
# before doing any distance-base filtering, we use tap_lines to trip maz-tap pairs to the nearest tap served by tap_line
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 652057 to 109996 rows based on tap_lines
 
# it looks like there are only taps > 3 miles if the nearest tap is more than 3 miles away - sometime it can be quite far
activitysim.core.los - maz_to_tap table maz_tap_walk.csv max WALK_TRANSIT_DIST 14.64564393939394
 
### max_dist: 0 (i.e. no max_dist cutoff)
 
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 652057 to 109996 rows based on tap_lines
activitysim.core.los - maz_to_tap table maz_tap_walk.csv max WALK_TRANSIT_DIST 14.64564393939394
 
activitysim.core.tracing - Time to execute run_sub_simulations step mp_tvpb : 670.81 seconds (11.2 minutes)
activitysim.core.tracing - Time to execute run_sub_simulations step mp_mode_choice : 346.731 seconds (5.8 minutes)
INFO - Time to execute all models : 1394.708 seconds (23.2 minutes)
 
### max_dist: 3
 
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 652057 to 109996 rows based on tap_lines
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 109996 to 102059 rows based on max_dist 3
 
activitysim.core.tracing - Time to execute run_sub_simulations step mp_tvpb : 666.831 seconds (11.1 minutes)
activitysim.core.tracing - Time to execute run_sub_simulations step mp_mode_choice : 315.394 seconds (5.3 minutes)
INFO - Time to execute all models : 1299.45 seconds (21.7 minutes)
 
### max_dist: 2

activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 652057 to 109996 rows based on tap_lines
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 109996 to 66661 rows based on max_dist 2

activitysim.core.tracing - Time to execute run_sub_simulations step mp_tvpb : 675.95 seconds (11.3 minutes)
activitysim.core.tracing - Time to execute run_sub_simulations step mp_mode_choice : 184.012 seconds (3.1 minutes)
INFO - Time to execute all models : 1227.788 seconds (20.5 minutes)
 
max_dist: 1:
 
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 652057 to 109996 rows
activitysim.core.los - trimmed maz_to_tap table maz_tap_walk.csv from 109996 to 32048 rows based on max_dist 1
 
activitysim.core.tracing - Time to execute run_sub_simulations step mp_tvpb : 672.937 seconds (11.2 minutes)
activitysim.core.tracing - Time to execute run_sub_simulations step mp_mode_choice : 99.481 seconds (1.7 minutes)
INFO - Time to execute all models : 1077.644 seconds (18.0 minutes)

6a Task 2 Complete Estimation Mode (milestone)

Doyle now getting started on Newman's couple of items + refactoring the trip models

6a Task 3 Telecommute Model in Cooperation with SEMCOG (milestone)

Updated the Telecommute design page based on our discussion last time
Plan to commit the draft example setup for programming to the repo for everyone to see as well

6a Task 8 Maintenance and Support (milestone)

Discussed plan earlier this week for reconciling the various pull requests in order to release an updated version
Will release a new version before end of 2020 for the 5b work (and some of the 6a work too)
PSRC testing code and will try using the multizone code (and dynamic chunking capabilities)
SFCTA is going to work on an activitysim setup and this could replace the toy 2 zone example that we have
PSRC has a couple support questions that they will email us about
We really need to setup a roadmap on the wiki that includes all work being done by this project, the ARC project, and the SEMCOG project
This roadmap will become even more important once the bench is setup
I'll create a roadmap wiki page to seed this. We'll include a 6 month look out
It's a bit difficult for us to say with any certainty when features will be contributed and then released from other projects since we don't have control over them
Just to confirm, ARC will make MTC data examples for the new features they are adding
Its important we be a good gatekeeper w.r.t. design, software, examples, documentation, etc.
The SEMCOG telecommute exercise is a good example of making cooperation an explicit part of our scope
If there's no coordination with the contributor then there's a good chance the contribution will get turned down
New big features, such as transit capacity, crowding, and reliability need a switch to turn off or on depending on interest
This is an important topic that we need to continue to discuss