allow denser packing of stopped bicycles

[matkoniecz]

Currently bicycle lane takes as much pace as car/bus lane what is understandable (for simplicity reasons, though you can remove one car lane and have enough space for two bicycle lanes).

But it results in very weird results in case of bicycle traffic jams, with extremely low density of stopped bicycles - what increases risk of deadlock.

Would be feasible to allow smaller gaps or make bicycles smaller? I guess that two lanes of bicycles in one lane would be obnoxious to implement.

Compare

[dabreegster]

I'll think about this.

One idea is having variable lane widths. Even a parking lane and a regular vehicle lane aren't the same width in reality. There are implementation difficulties to editing lane types with different widths, but they can be overcome. More importantly, we're taking wild guesses for the widths of roads today. If we initially guess there is or isn't room for some edit, it's easy to get it wrong. Treating all lanes as the same width plays it safe in some cases, but not in others.

The other idea is treating every bike lane as having double width. Kind of related to #139. Maybe there should be different behavior for waiting at the end.

[matkoniecz]

One idea is having variable lane widths. Even a parking lane and a regular vehicle lane aren't the same width in reality.

That also would be great (and thought about opening issue for that), though even on standard bicycle lane (for one bicycle in width, a bit less than half of car lane width) there should be still smaller distance between stopped bicycles.

[dabreegster]

@Robinlovelace reported the same thing again. Worth revisiting this now that we have variable lane width and a start to dynamic lane-changing.

First question: Should this denser packing behavior only happen when stopped? From my US-centric experiences, strangers don't ride this close going at speed except in critical mass. If so, any thoughts how the "unfurling" behavior should look? When everyone starts going again, there might be some natural variation in speed (especially up inclines), so should people mostly just fall back into a single queue?

I have some old ideas how to implement this (https://github.com/a-b-street/abstreet/issues/139#issuecomment-654421970) that I've never tried. Given the new work in dynamic lane-changing, might revisit this or not.

[Robinlovelace]

Thanks for the dedupe @dabreegster and apologies, I did a search of issues containing the word 'queuing' but somehow didn't see this. In answer to your questions, my 10 cents are:

Should this denser packing behavior only happen when stopped?

Partially yes. I think the bikes could do with being denser when moving, e.g. 2 or 3 abreast, but denser still when parked, e.g. 4 to 6 abreast.

From my US-centric experiences, strangers don't ride this close going at speed except in critical mass. If so, any thoughts how the "unfurling" behavior should look?

Woa hadn't thought about unfurling. Maybe a randomised 'set-off from parked' time?

When everyone starts going again, there might be some natural variation in speed (especially up inclines), so should people mostly just fall back into a single queue?

Not 100% sure what you mean by this but sounds reasonable and the variation in speed could mean that unfurling happens enough anyway for the short section of time when it is important. Definitely an important one for modelling scenarios of high cycling uptake, as done in the PCT, so very glad you're looking into it, many thanks!

[matkoniecz]

Should this denser packing behavior only happen when stopped?

Cyclists are packed more densely than cars also when moving, but not as packed as when moving.

In particular two cyclists next to each other on full car lane is easy to fit (but would not be done often in city due to traffic).

But also distance between moving cyclist is smaller than between cars. https://user-images.githubusercontent.com/899988/87248960-8ffd5100-c45c-11ea-86c7-b3f6108aeea3.jpg shows 3 moving cyclists in space taken by two stopped cars, and probably far less than one moving car.

[wookey]

Sumo has been through all this in the past so we might be able to learn things from them. Has anyone looked? It started with the same model of strict lane use but I understand got forked for Asian cities with lots of cycles and mopeds where the 'strict lane use' model was wildly inaccurate. I've not looked at that software since 2018 so I don't know if this stuff got merged back in to the main project.

[dabreegster]

SUMO calls it a sub-lane model: https://sumo.dlr.de/docs/Simulation/SublaneModel.html

This issue is about stopping distances, but I also have old ideas about modeling shared bike/pedestrian spaces with simpler over-taking. I think of it like an accordian: by default, people mostly stick to the left or right. When a bike wants to overtake a pedestrian, they'll shift over, but return to the edge afterwards. These queues are bidirectional and not really divided into individual lanes, though they might have some kind of width (max 3 people or bikes at a time). I had some notes long ago about how to make this work efficiently with discrete events, but never pursued the idea. Maybe a good plane ride project.

[Robinlovelace]

Is this issue also for discussing packing of moving bicycles? As discussed in this thread the the packing is related with speed but I'm sure more packing at all speeds would be a good thing when visualising scenarios of high cycling uptake like Go Dutch in the PCT with A/B Street.

[matkoniecz]

I intended this issue to be solely about stopped ones as effect there are extremely blatant (10-20-30 times lowered density). While for moving ones it is on edge of expected, maybe should be increased but by factor of 2 maybe 3 at most.