universe-cuda container: [ERROR] [launch_ros.actions.load_composable_nodes]: Failed to load node 'occupancy_grid_map_node'

[rt-2pm2]

I am having some issues in trying to run the Autoware framework from the universe-cuda container:

REPOSITORY                            TAG             IMAGE ID       CREATED      SIZE
ghcr.io/autowarefoundation/autoware   universe-cuda   714e9b03ca81   3 days ago   12.5GB

Precisely, trying to run the e2e_simulator.launch.xml launch file, there are some issues that I spotted in the logs (by filtering for ERROR):

[ERROR] [launch_ros.actions.load_composable_nodes]: Failed to load node 'occupancy_grid_map_node' of type 'autoware::occupancy_grid_map::PointcloudBasedOccupancyGridMapNode' in container 'pointcloud_container': Failed to find class with the requested plugin name.
[component_container_mt-1] [ERROR] [1747066055.358206833] [pointcloud_container]: Failed to find class with the requested plugin name 'autoware::occupancy_grid_map::PointcloudBasedOccupancyGridMapNode' in the loaded library
[component_container_mt-57] [ERROR] [1747066056.639504428] [perception.traffic_light_recognition.traffic_light.classification.car_traffic_light_classifier]: please install CUDA, CUDNN and TensorRT to use cnn classifier
[component_container_mt-57] [ERROR] [1747066063.979023157] [perception.traffic_light_recognition.traffic_light.classification.pedestrian_traffic_light_classifier]: please install CUDA, CUDNN and TensorRT to use cnn classifier

I don't see those errors when launching the same launch file on the host native, using the "from_source" Autoware.
I pulled the last version of the container to be sure to have the updated version.

I am able to run the same demo on the same machine, but leveraging the "from source" installation of the Autoware Framework.
After launching the container with:

<autoware_src_path>/docker/run.sh --map-path <host path to the autoware_map folder> --data-path <host path to the autoware_data folder> bash

where I slightly modified the run.sh to add a custom ROS_DOMAIN_ID, which translate into:

docker run -it --rm --net=host --gpus all -e LOCAL_UID=1000 -e LOCAL_GID=1000 -e LOCAL_USER=<myuser> -e LOCAL_GROUP=<myuser>  -e DISPLAY=<myhostname>:10.0 -v /tmp/.X11-unix/:/tmp/.X11-unix -e XAUTHORITY= -e XDG_RUNTIME_DIR=/run/user/1000 -e NVIDIA_DRIVER_CAPABILITIES=all -e TZ=XXX/YYY -e ROS_DOMAIN_ID=<MY_ROS_DOMAIN_ID> -v /home/<myuser>/autoware_map:/autoware_map:ro -v /home/<myuser>/autoware_data:/autoware_data:rw ghcr.io/autowarefoundation/autoware:universe-cuda /bin/bash

I am running the demo in the container with:

ros2 launch autoware_launch e2e_simulator.launch.xml map_path:=/autoware_map/nishishinjuku_autoware_map vehicle_model:=sample_vehicle sensor_model:=awsim_sensor_kit launch_vehicle_interface:=true

Are you aware of any issues with the universe-cuda runtime container? Can someone confirm that the e2e_simulator demo can be run on that container?

Notes

The Autoware source is updated up to the following commit:

commit 886277580b27f9bfa60f2cb5c4e7d4d687211c46 (HEAD -> main, origin/main, origin/humble, origin/HEAD)
Author: Yutaka Kondo <yutaka.kondo@youtalk.jp>
Date:   Fri May 9 17:59:41 2025 +0900

[mohamedleithy]

Any updates on this issue?