Handling Bikeways

[mehmetdogru]

🚲 In the YTU Davustpasa Campus where we are currently testing our self-driving vehicle, we have some scenarios which ego vehicle have to be on the bikeways to be able to drive. To keep the continuity of autonomous drive and to assure the safety of the bikes at the same time, we are thinking of implementing a module in behavior_path_planner (Details TBD) Please feel free to share your ideas on this topic. I will update this discussion soon with more details.

[mehmetdogru]

cc: @TakaHoribe @xmfcx @mitsudome-r @maxime-clem @takayuki5168

[maxime-clem]

I think the dynamic drivable area expansion can work for the YTU Davustpasa Campus as projecting the ego vehicle along the centerline would expand the drivable area into the bike lane. I will build a simple scenario to test it.
However I can understand the concern about performance and the proposed solution also sounds good.

[mehmetdogru]

Thanks @maxime-clem. I think we will complete this obvious overlapping road-bikeway part of the map today and I will share it here if you would like to use it to test.

cc: @mozhoku

[mehmetdogru]

@maxime-clem please find the map files on the link. I marked the related part as follows however bicycle_lanes are not visible on rviz yet and the annotation might not be very right at the moment since it is the first trial. Hope this helps. Thank you.

[maxime-clem]

Thank you and sorry for the late reply.
The dynamic drivable area expansion works well for this case and can be considered as a temporary workaround.

dyn_da_expansion.bicycle_lane.mp4

[xmfcx]

• https://sfbike.org/news/bike-lanes-and-right-turns/
  • 
• https://nacto.org/publication/urban-bikeway-design-guide/intersection-treatments/through-bike-lanes/
  • 
  • 
• http://www.bikeirvine.org/news/2023/2/16/safety-at-intersections
  • 
• https://www.mcleishorlando.com/insights/drivers-or-cyclists-who-has-the-right-of-way/
  • 
• https://www.trthaber.com/haber/guncel/konyada-ogrenciler-okullarina-guvenli-bisiklet-yolu-ile-gidiyor-681609.html
  • 
• https://www.youtube.com/watch?v=T_VUXUGjrUw
  • 

[mehmetdogru]

What if we annotate bicycle_lane and shared_bicycle_lane separately and so in cases like below we would be including the neighbor shared_bicycle_lane into account. Also with that approach we can manipulate the path easier so we would be using the existing modules as well. Plus we can support scenarios like above as well which @xmfcx has shared where there are shared bicycle lanes for routing purposes

What do you think?

cc: @maxime-clem @xmfcx @TakaHoribe

PS:

• The most recent version of the map with the bicycle_lanes & shared_bicycle_lanes are on this link.
• You can check out these branches to visualize bicycle_lanes & shared_bicycle_lanes
  • https://github.com/mehmetdogru/autoware.universe/tree/feat/visualize-bicycle-lanes
  • https://github.com/mehmetdogru/autoware_common/tree/feat/add-bicycle-lanelet-query

[maxime-clem]

This representation looks good.

Maybe I forgot, but why don't we simply represent the shared space (shared_bicycle_lane) as a normal lanelet where the autonomous vehicle can drive ? We can keep the centerline of the current lanelet, but move the right bound to be the right bound of the shared_bicycle_lane.

[mehmetdogru]

Thanks @maxime-clem

In such cases bicycles on shared_bicycle_lane has the priority so to be able to satisfy this and so ensure the safety of the bicycles I think it will be good to annotate it.