[Autoware.Auto] euclidean cluster node detects surroundings as huge bounding box

asked 2022-01-07 03:34:31 -0500

TopRamen gravatar image

Hi everyone!

I’m in the process of setting up Autowara Auto on our robot. I managed to set up the autoware stack on it an when testing i have encountered issue with obstacle detection. I can't quite get the euclidean_cluster node to adapt to a more enclosed environment in which we are testing. It takes the whole space as one large obstacle as if the robot is inside it (as you can see here). I have tried playing with parameters in euclidean_cluster.param.yaml but didn't manage to limit the object size.

Is there a way to adapt the euclidean cluster node to work in a more enclosed environment?

I should also add that I'm working with a up to date version of Autoware.Auto stack sourced from /opt/AutowareAuto in ade container.

edit retag flag offensive close merge delete

Comments

I'm working with a up to date version of Autoware.Auto stack sourced from /opt/AutowareAuto

You can try to build the latest version of the code (instructions here). I am not sure if this will help.

You can try to completely deactivate obstacles by launching Autoware with the argument with_obstacles:=false (e.g., ros2 launch autoware_demos avp_sim.launch.py with_obstacles:=false).

Maxime gravatar image Maxime  ( 2022-01-11 00:56:38 -0500 )edit

By the way, here is the Gitlab issue related to this problem: https://gitlab.com/autowarefoundation...

Maxime gravatar image Maxime  ( 2022-01-18 16:54:43 -0500 )edit
1

Hi, sorry for my late reply. I'm aware of the issue on gitlab, I was commenting there allso to explain the issue. When it is solved i will link the answer here.

TopRamen gravatar image TopRamen  ( 2022-01-24 03:07:23 -0500 )edit