dabreegster
commented
3 years ago Poundbury
Bridport road and Damers are the bottleneck into/out of this site. These cars and the cyclist all want to go to Dorchester, but there are only two direct routes there. The roads have a single lane each direction, with barely any shoulder, from checking street imagery. In abst, there's no lane over-taking modeled yet, so the cars stack up behind the cyclist, making the bottleneck very apparent to the player. In reality, either many cyclists would take a much more circuitous route, they'd attempt to use the tiny sidewalk, or cars would be passing bikes in a manner that's unpleasant for both.
According to street imagery, Bridport has intermittent street parking starting around Williams. If we're focusing on this site particularly, it'd be worth tagging it.
One edit the player could attempt here is making Bridport and Damers one-way for cars, using the other lane for bikes. In reality, one car lane could become a two-way bike lane, but as mentioned at the top of this issue, that's not possible in abst yet. So here's a rendition of that, with Bridport just eastbound into Dorchester and Damers westbound into Poundbury:
oneways.json.txt
This one-wayness ends at Albert Rd / Cornwall in Dorchester, although maybe it'd be better to extend it all the way to South St. Here's the baseline simulation with this change:
Green is good (cyclists using dedicated lanes of some sort), red means cars & bikes sharing the same lane. But note red isn't necessarily bad; this view doesn't show if a bike and car were actually conflicting at that moment. Also note Damers is unused, because our scenarios still need return trips.
For comparison, here's the same view without any edits:
Hovering over each road, about half the cycle traffic actually wants to use Damers, because it's a more direct route.
Some short-term TODOs to tell this story better:
- include return trips in the scenario
- measure cyclist/driver conflict events -- every time a car is immediately queued behind a bike. Count it per road segment. In reality, overtaking would likely occur here (and will be modeled that way in the future).
- Record the simulation results for baseline and go active, so easy comparisons can be made with any map edits. Have to first decide what sort of background traffic to include.
Next up, Leeds -- Robin showed me a specific idea around Headrow and Vicar.
Robinlovelace
I'll use this issue to record observations about how individual sites look in A/B Street, with the aim of generating material to focus on in the workshop presentation. My general goal is to establish how abst adds value to the actdev story beyond the 2nd zoom level.
What I'd expect: abst's value is in the feedback loop of observing some problem in the simulation, editing the map to intervene, comparing quantitative results and watching the change, and repeating. The biggest challenge currently is that roads can't be widened, so if OSM doesn't have street parking or multiple lanes tagged in the sites, then the player has less opportunity to add cyclepaths.