(I'm terrible at sticking to one goal -- haven't been focusing on data quality in a bit, so this is out-of-date.)
How usable is each map overall? Different milestones (roughly in order):
- doesn't crash, but gets permanently stuck in gridlock
- full PSRC day finishes without gridlock
- with freeform policy is usually easier, since traffic signal defaults are quite bad
- full day finishes without any aborted trips
- enough parking spaces exist
- parked cars are available when trips need them
- no disconnected paths
- no geometry problems; map looks reasonble and represents reality
- simulation results match available measurements of reality
- known vehicle counts for roads
- example problem: lots of traffic cutting through UW
The different map slices are pretty arbitrary, but roughly get bigger down the list.
- full PSRC day finishes with traffic signals
- ~300 aborted trips, no free car available
- full PSRC day doesn't gridlock with freeform policy
- 23rd and madison traffic signal is the biggest problem, try manual traffic signals
- short roads at: 12th/union/madison, boylston/roanoke
- not enough room to even seed parked cars!
- uninvestigated crash a while in
- short roads: aurora/winoa
- uninvestigated crash a while in
- super weird lane geometry near 26th and gilman
- redundant points in a straight line
- short roads: lake city, ne 115th
- 12th and ravenna seems stuck, not sure why
- Car with one-step route Router, crash around 6am