A possible bug in scenario_selector when multiple parking lots are available?

[kminoda]

I tried a parking scenario in Autoware, but it behaves oddly when there exist multiple parking lots on a map.

I hope you can get the point with the following video:
https://user-images.githubusercontent.com/44218668/166207450-2950ad13-2281-4e24-abee-ccaeef6625cb.mp4

From now on, I will refer to parking_space as a parking spot and parking_lot as a parking area (the same way as defined in lanelet2).

Briefly looking at the implementation of scenario_selector, it appears that it proceeds to the parking scenario when the following two conditions are met:

1. the goal is on a parking_space
2. the vehicle is on the nearest parking_lot

However, when multiple parking_lot exist, the vehicle starts parking even though it's on a different parking_lot that corresponds to its goal parking_space.

I think the most simple solution would be to search for the parking_lot that its goal parking_scape belongs to. What do you think?

[kenji-miyake]

Actually, the parking scenario isn't tested in many use cases.
With the current implementation, it simply combines all cost maps from lanes, parking lots, and parking spaces, and finds a path.

You can show an ideal use case scenario and fix the code to accomplish the use case.
In this situation, I feel it's good to start with the lane driving scenario.
So I guess you should fix the initialization condition to check if it's on a lane.
Or as you say, you can check there if the vehicle is in a parking lot of the target parking space.

[kminoda]

OK, thanks for the suggestion!
I will fix the bug, and if successful, I will open an issue or PR.