Many service roads and driveways were previously excluded from osm2streets entirely. Undo that. The work on block tracing and junction/corridor bundling needs to take those into account.
And get rid of an old transformation that tried to trim very short dead-end roads. I'm pretty sure the motivation for this was traffic sim related -- U-turns on those very short dead-ends caused a really funny (in hindsight...) gridlock case, where a vehicle blocked itself.
In the spirit of getting closer to a faithful representation of OSM data, gut both of these out. Some more similar cleanups may follow.
I visually inspected all test diffs, and they look fine. Some areas get more complex with driveways, but that's real complexity we want to see and work on.
Many service roads and driveways were previously excluded from osm2streets entirely. Undo that. The work on block tracing and junction/corridor bundling needs to take those into account.
And get rid of an old transformation that tried to trim very short dead-end roads. I'm pretty sure the motivation for this was traffic sim related -- U-turns on those very short dead-ends caused a really funny (in hindsight...) gridlock case, where a vehicle blocked itself.
In the spirit of getting closer to a faithful representation of OSM data, gut both of these out. Some more similar cleanups may follow.
I visually inspected all test diffs, and they look fine. Some areas get more complex with driveways, but that's real complexity we want to see and work on.