Rubidium wrote:

The pathfinder just thinks that the waiting takes shorter than the driving around it.

Furthermore making it go to the other side causes the same:
a) one vehicle at the station as it is now and then four vehicles are going to the stop. They see one side is empty and all go to the other side.
b) same as the above case, but when a slot is freed/used all other vehicles going to the station rethink the side to go to. This can cause a vehicle indefinitely do go around the station without ever having the chance to load.

So as it stands there are no viable issues that do solve the issue and do not cause unwanted side effects.

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944#comment3970

Expresso wrote:

Well, maybe it's a good idea to implement a system in which road vehicles reserve a platform and wait for that platform to come free?
Of course a road vehicle should only try to reserve a platform if it's at a certain distance from the station (less then 10 tiles, for example). If all road stops are full, the road vehicle should be put on a list of waiting vehicles (fifo) to prevent strange situations and either drive in circles for a while or stop somewhere (at this point the traffic lights patch could come in handy).

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944#comment4620

frosch wrote:

Just to mention it:
There is another aspect of this bug in nightlies. The slot reservation only works for road stops which are part of the orders. But now road vehicles can also stop at other road stops (non-non-stop orders).

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944#comment5100

frosch wrote:

Note: YAPF performed especially bad with drive-through stops. Since r14703 it is as good/bas as NPF at least.

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944#comment5128

Expresso wrote:

The pathfinder (yapf at least) seems to treat drive-through road stops the same as normal road stops. If (for example) there's a north-south DT road stop, a vehicle arriving from the south should notice that the other platform can only be accessed from the north. CYapfCostRoadT.OneTileCost (lines 63-66 in yapf_road.cpp) doesn't seem to check for this.

Could this be the culprit?

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944#comment6164

frosch wrote:

The problem consists of at least four parts:

1. The roadstops cannot dinstinguish which slot is reserved by a road vehicle, resp. how many vehicles are heading for which slot.
2. The multistop code cannot compute the costs for reaching the different slots, as all pathfinders only get a destination tile. Same applies to the vehicle itself after reserving a slot. dest_tile cannot identify the direction.
3. Driving through roadstops of a station to reach another roadstop of the same station won't work. (Again applies for both pathfinding during reservation and during driving)
4. When multiple drivethrough stops are sequential in a row, they would need to be treated as one drivethrough stop. But depending on long vehicles block multistop station #1495 with less penalty for multiple vehicles heading to it.

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944#comment6167

Rubidium closed the ticket.

Reason for closing: Fixed

In r18404; might need a higher occupancy penalty before it actually thinks driving around is better; primarily because the stop is half filled.

This comment was imported from FlySpray: https://bugs.openttd.org/task/1944