2026-08-25 Australia Team

[joecastiglione]

Agenda

• Admin
  • Governance Team
    
  • Task Order Amendment Policy
    • CS Task Order Amendment
    • Outer Loop Task Order
  • RFQ Pre-submittal Meetings
• Phase 11 Tasks
  • MWCOG Telecommute Frequency Sensitivity Test
    • Results / Next Steps
    • Scope
  • Application and Analysis Guide
    • Current Draft
    • Scope
• Other
• Ancillary model inclusion
  • eg. SANDAG Airport Model
• Technical
  • Shadow pricing / simulation constraint issues
    • Confirmation and validation of methods
    • Investigation of results
    • Usability improvements and documentation
  • Parquet Support Testing
  • Adaptive Chunking Support
    • Issue
    • Discussion
• Phase 12 Issue Template

Notes

Governance Team and Task Order Amendment Policy

• At the direction of the Executive Team, Joe formed a governance team (one voting rep per participating agency, 14 of 15 agencies currently represented) to formalize administrative decisions as the project continues to grow in complexity and scale.
• Decisions are posted as GitHub issues in a governance repo, with agencies voting.
• New task order amendment tiers: under $5,000 = executive committee decides, full PMC advance notification not required; $5,000-$15,000 = executive committee decides, full PMC advance notification required; over $15,000 = full PMC approval required.
• Recent examples: a sub-$5,000 amendment for CS's Application and Analysis Guide budget, and a ~$9,900 new task order to Outer Loop for Jan to co-lead engineering (Jeff Newman will miss about half of meetings for the much of the rest of the year).
• ACTION (Paris): confirm DTP's governance rep (likely him) and send username to Joe.
• ACTION (Amir): share the GitVote GitHub app link as a possible cleaner voting tool.
• ACTION (Joe): compare GitVote against the governance document to ensure all voting needs are met; add the missing task order link to the tracking doc.

RFQ Per-Submittal Meetings for Consultant Bench Selection

• Two pre-submittal meetings held last week; 6-7 consultancies attended
• Written questions due tomorrow; responses/addenda post within a week; qualifications due about two weeks after.
• VLC and KPMG inquired about the template contract for legal review; VLC's counsel did not identify any barriers, KPMG has not responded.
• ACTION (Joe): Discuss with / email agencies next week requesting agencies to identify staff to participate in scoring proposals.

Techincal Updates

• Telecommute frequency sensitivity testing (MWCOG): A borrowed TransLink mode-choice logsum parameter produced greater sensitivity to cordon pricing (vs. almost none before), but shifted the model's calibration significantly. Next steps (recalibration vs. re-estimation) still undetermined.
  • ACTION (MWCOG/RSG/Joe): hold email/follow-up discussion to define next steps; Joe to share results.
• Application and Analysis Guide
  • Draft received from Joe Flood/CS, accessible in the CS repo.
  • ACTION (All): review draft and give feedback within the next couple weeks (goal: wrap up all Phase 11 work by end of September).
  • ACTION (Joe): send formal email pointing people to the draft.
• Adaptive chunking
  • An external contributor (TU Delft) submitted pull requests to fix adaptive chunking for memory-constrained/Kubernetes use
  • Adaptive chunking is little-used in implementations (versus explicit chunking)
  • Has known robustness issues
  • Open question: invest, deprecate, or leave as-is without a formal recommendation.
  • ACTION (Jan/Amir/Joe): continue discussion with the Product and Community Team on chunking.
• Parquet support testing (Amir/Jan)
  • Full-skim tests show a higher peak memory spike than OMX but ~5 min faster skim-reading (of ~7 min total)
  • Jeff plans further optimization
  • Group agreed to support both formats.
• SANDAG simulation-based location choice
  • SANDAG found anomalies (zones with many workers but no chosen destinations; oversubscribed zones) after ~10 iterations
  • Investigating cause
  • DTP (Paris) previously moved away from simulation-based location choice due to unrealistic long trips from residual jobs; open to revisiting once model reaches production status.
  • TfNSW sees no such anomalies (uses occupation-based worker segmentation, no double-constraining by segment).
  • ACTION (Amir/TfNSW): consider sharing tail-calibration/config methods with the group via a call or GitHub issue.
  • ACTION (Joe): raise with engineering whether configurable thresholds could mitigate unrealistic long-distance trip assignments.

Other Items

• Ancillary model scope
  • Question raised (via a code cleanup) about whether models outside core passenger travel demand (e.g., an airport model) belong in ActivitySim's scope
  • Unresolved in the roadmap - flagged for Phase 12.
• Phase 12 issue template
  • Joe's request for feedback on the template
  • Meant to standardize issue detail for Phase 12 prioritization.
  • ACTION (All): review and mark up the template.
  • ACTION (Joe): consider running with the current version and gathering feedback in about three months (per Shaun's suggestion).

SUMMARY OF ACTION ITEMS

• Paris - confirm DTP governance rep, send GitHub handle (done)
• Amir - share GitVote link with Joe.
• Joe - compare GitVote to governance document; add task order link to tracking doc.
• Joe - email agencies inviting RFQ scoring participants.
• Joe/MWCOG/RSG - determine next steps on telecommute model recalibration.
• All - review and comment on Application and Analysis Guide draft.
• Jan/Amir/Joe - continue adaptive chunking discussion with Product and Community Team.
• Amir - share tail-calibration/config methods for simulation-based location choice.
• Joe - raise threshold-based mitigation options with the engineering team.
• All - review and mark up the Phase 12 issue template.