psrc / psrc_activitysim

PSRC.org Website

Big picture

• Currently running rails v2.0.2.
• CMS built on Radiant CMS.
• Deploys to BlueBox using capistrano.

Notes

• In production the app runs with apache/passenger. Apache redirects can be defined in /sites/psrc-production/shared/htaccess. This file is symlinked to the public dir (which is the apache root) on cap deploy.

psrc_activitysim's Issues

change mp settings to allow user to specify number of procs for accessibility

Implement Telecommute Model

Borrow from one of the examples.

Destination Density variables in Tour Mode Choice using origin Density

For example:
https://github.com/psrc/psrc_activitysim/blob/main/configs_dev/tour_mode_choice.csv#L60

But destination density does exist, so you use that instead:
https://github.com/psrc/psrc_activitysim/blob/main/configs_dev/tour_mode_choice_annotate_choosers_preprocessor.csv#L22

Check density calcs in tour mode choice pre-processor

Map them and see if they look ok.

Trip scheduling has lots of fails

Pretty sure this is a known issue and has a solution.

Add school bus as a tour/trip mode

We should account for these even though we don't assign them.

Fix home_is_rural

ActivitySim/activitysim#658

'start' and 'end' seem to be same hour for tour mode choice estimation data bundle for all records.

Check the survey conversion?

Use only one mode alternative for TNC

Collapse single, shared and taxi into one TNC mode.

Auto ownership should be run before school/work location

School and work location uses mode choice logsums. Mode choice spec has auto availability variables.

Do not include 999 values for access_dist_transit calc

Some parcels within a maz may not have access to transit (999) so we should ignore these in the mean calc if others do have access.

Adjust time periods to match skims

Rename Auto modes

Change DRIVEALONE, SHARED2FREE & SHARED3FREE TO DRIVEALONE, SHARED2 & SHARED3.

Review estimation data filtering

Compare to SANDAG when available.

Recent log for 3-day sample:

02/08/2023 04:50:05 PM Dropped 525 trips with purpose 'Change mode'
02/08/2023 04:50:05 PM Dropped 4 trips with null purpose
02/08/2023 04:50:05 PM Dropped 2837 trips with same origin and destination MAZ
02/08/2023 04:50:09 PM Dropped 1934 trips because person was removed from data
02/08/2023 05:14:35 PM Dropped 5 tours due to -1 primary purpose
02/08/2023 05:18:56 PM Dropped 518 tours for too many stops
02/08/2023 05:18:56 PM Dropped 2008 persons: duplicate IDs from canonical trip num definition
02/08/2023 05:18:56 PM Dropped 3 persons: missing an outbound or inbound trip leg
02/08/2023 05:19:06 PM Dropped 7 tours: ptype 8 making work tours
02/08/2023 05:19:06 PM Dropped 0 persons: missing household records
02/08/2023 05:19:07 PM Dropped 0 tours: missing tour type
02/08/2023 05:19:07 PM Dropped 255 tours: missing tour mode
02/08/2023 05:19:07 PM Dropped 118 tours: parent tour was removed/missing
02/08/2023 05:19:08 PM Dropped 488 tours: missing inbound or outbound trip component
02/08/2023 05:19:22 PM Dropped 94 tours: travel day passes beyond midnight
02/08/2023 05:19:23 PM Dropped 1852 tours: missing inbound or outbound trip component
02/08/2023 05:21:02 PM Dropped 1038 tours: first half tour destination does not match last half tour origin
02/08/2023 05:21:02 PM Dropped 4 tours: under 16 making work tours

Create P&R Fare skims

Currently we are using walk to transit skims. Skim using the P&R assignment and skimming procedures.

Remove MAZ based transit walk access/egress in mode choice models (for now)

Specs include both AUX skims, which we use to represent transit walk times as well as the MAZ based access/egress times. So we are double counting. Remove MAZ based for now and use skims. Figure out a way to measure MAZ based distance/times for each sub-mode (ferry, lr, cr, bus) and also a way to represent transfer walk times as a matrix.

The rational for using skims (for now) is they actually measure the walk components for each sub mode, which can vary for each OD. The current data for MAZ access and egress distance/time would be constant across sub modes.

Calculate TOTACRE from MAZ polygon

Currently, the code is using parcel based area, but let's keep things simple and just use the polygon area from the MAZ file. This field is used often downstream in density measures.

Consider using tour is school bus in trip mode choice

Not implemented yet, but this is available via the trip mode preprocessor as 'tourSchBus'.

Figure out how 'DIST' skim is calculated

This is from the MTC implementation. Look for documentation and, if none, post issue on main ActivitySim repo.

NaNs in tour mode choice destination logsums

What causes this?

Columns mismatch in tour_mode_choice.csv

@tedlini In the dev_RSG branch, utilities for commuter rail & ferry were listed under the WALK_LR columns.:
https://github.com/psrc/psrc_activitysim/blob/a67df0f34a24ff0d25d5ed81689175570a6030cc/configs_dev/tour_mode_choice.csv

I fixed this with this commit:
8aab45c

But can you review to make sure my changes look good? Thanks!

Add new density_index calculation

This is a parcel buffer based variable that is aggregated to each MAZ.

Implement Disaggregate Accessibilities

Create auto skims segmented by VOT

Check tour/trip mode transit availability assumptions

Right now a cap on IVT is only set for the 2nd leg of the tour:
https://github.com/psrc/psrc_activitysim/blob/main/configs_dev/tour_mode_choice_annotate_choosers_preprocessor.csv#L64

Re-run size terms script to include FOOEMPN

Figure out Topology in land use

Right now we are assigning 1 to all MAZs. How has this been calculated in other implementations? I feel like we may have removed variables suing this field from some of our UECs. Put back?

Missing HH income in survey

Should we impute?

Make work from home sharrow compatable

@tedlini I am just moving the 'other_than' expressions to a preprocessor to make it work with sharrow. Similar to what SEMCOG did.

Completely remove topology variable from land_use file and all UEC's.

We are not using this variable.