We need to talk more with Jeff and Joel about the approach
There's some concern about inefficiencies in the approach and maybe we should revise the design a bit
We want a TVPB that considers various tradeoffs
behaviorally strong
built with good technology
manageable
affordable
makes use of reasonable input data
etc.
SANDAG has about 60 taps per maz as well
But not all maz-tap pairs are fully evaluated because some taps are further away and have duplicate service, some tap-tap pairs have no in-vehicle time, etc.
We pre-compute the path components and then select N best paths for omaz, dmaz, tod, demographic segment
It's important the user builds a reasonable maz-tap network or otherwise there are major performance issues
Need to sharpen up our focus, describe tradeoffs, unknowns, etc. We're spending a lot on this task and we need to make sure we have a good plan.
So let's pause for a sec and discuss the design some more
Another feature we could add is label taps by mode, sort by distance, and then make sure we get at least one tap for each mode up to a max distance
With new technology like e-bikes, max transit access distances get even longer so more paths considered
Next steps, let's get some documentation on TVPB, pros/cons, design considerations, how geography is defined, challenges (data, runtimes, etc.)
Assignment algorithms are getting better and better, can we just use them instead?
TVPB has lots of good features - builds path sets with more line-haul modes, solves issues with transit access connector issues in network assignment, more sensitive to policy issues, etc.
Do we have some proof of the benefits? Maybe a study with and without TVPB?
Bill, Wu, Joel, Jeff, me collect some info to discuss Tusday