Asking about effective governance to help guide and advice other government agencies. One objective is to extract lessons and best practices to be applied elsewhere. They have pointed to ActivitySim as a success, noting good, robust documentation.
They want to understand how our governance structure developed and adapted, how has it helped us to be successful.
What they would like from us is: confirm it is ok to be a case study for their research, to conduct interviews with consortium members, and to observe some of our meetings.
They have monthly meetings (usually first Thursday of every month), which we are welcome to join, or be apart of our newsletter. Information can be found on their website, or contact Colleen directly.
Joe: Our governance document is something that we have but don’t follow strictly, probably worth revisiting. Our governance is more than the document but also how we run our meetings and other things that we are doing now.
Alex: Champions is also an interesting aspect to look at for how successful these collaboratives are - who’s engaged, who’s running the ship, and how effective they are. Also suggests exposing weaknesses (or opportunity areas for growth).
Mutual benefits – ActivitySim could learn from what others are doing.
Intergovernmental software collaborative to share drafts/findings as they are produced, we can be as engaged as we want to be in the review of those documents.
Next Steps: Joe to follow-up with Colleen
Other Admin
Caitlin noted that the 60-day reminder for invoicing goes out today. Agencies have 30 more days to pay. Please follow up with your departments if you haven’t paid yet. If you are unsure, please check with Caitlin.
AMPO conference confirmed attendees include: Jeff, Joe, and Dennis
Conference is the last week of October (session is October 26).
Joel can attend and volunteers to be on the panel.
MWCOG is planning to go (Feng Xie and maybe Mark Moran). Mark to provide email addresses of attendees to Caitlin and confirm is someone can be included in the panel.
Current implementation simulates trip schedules using probability distributions based on tour departure/arrival times.
SANDAG models implements a different methodology – applying probably distributions based on last known (tour or trip) departure or arrival time.
As the model simulates, it can "fail" by hitting a max number while trying to find logical trip departure/arrival times. Failure rate can get high (CMAP testing is about 20-25%). Unknown run-time impacts of simulation fails.
You could find a way to force the data to fit the algorithm, but not ideal. Algorithm should fit the data.
Alternative approaches
SANDAG resident model trip departure time: uses indexing to last known departure trip period rather than tour-based. Could still have inconsistent tours but fixes issues by half-tour.
SANDAG visitor and cross-boarder models: probabilities based on remaining time on the tour, expressed as a range and look up establishes offset for time. Can’t fail in either direction because of the structure and maximizes information by data.
ARC has a logit-trip scheduling model that is already implemented in ActivitySim
Potential next steps
Understand scope of problem quantitatively: where/how much is the simulation failing and run-time implications
Understand impacts, pros and cons, of implementing different approaches
Discussion
Defining time distributions without purpose streamlines approach but purpose is important for activity duration information
Group is unsure how to proceed, will think about it and revisit in a couple of weeks (Michelle added to Sept 15 agenda)