BARN.mp4
This is the repository for the proposed solution of Team MLDA_EEE (Machine Learning & Data Analytics) from Nanyang Technological University, Singapore
The original readme file is in README_BARN.md
Launch file at ./jackal_helper/launch/move_base_mlda.launch
Source files at ./mlda_algo
run_rviz_auto_start.py
: run thempc_node.py
and the nav stack with rvizrun_rviz_no_collision_nor_time.py
: run the environment only, use the 2D pose in rviz to set arbitrary goal to test the nav stack
We use Go 1.20 and Singularity 4.0.2 Help with Singularity
# Build image name 'nav_competition_image.sif'
sudo singularity build --notest nav_competition_image.sif Singularityfile.def
# Run
./singularity_run.sh ./nav_competition_image.sif python3 run.py --world_idx 0
- On Ubuntu 18.04 Machine
We can build and run the Singularity image. Recommended to build this with 18.04 and 20.04 machines
- On Ubuntu 22.04 Machine
We can build the Singularity image but it cannot execute the run.py
program through it due to GLIBC=2.34 missing
error and we cannot fix. In the .def
file, we also installed all the necessary ros-melodic-*
packages instead of relying on rosdep
This is the docker image for the April 1st Soft-Deadline submission
docker pull mldarobotics/barn2024:april1
Start the docker container named barn
in the background. Use VSCode Dev Container
extension to connect to the running container
# Allow GUI to be displayed
xhost +
# No Nvidia Container
docker run --rm -dt --name barn \
-e DISPLAY=$DISPLAY \
-e QT_X11_NO_MITSHM=1 \
-v /tmp/.X11-unix:/tmp/.X11-unix \
mldarobotics/barn2024:april1
# Nvidia Container
docker run --rm -dt --name barn \
--gpus all \
-e DISPLAY="$DISPLAY" \
-e QT_X11_NO_MITSHM=1 \
-e LIBGL_ALWAYS_SOFTWARE=1 \
-e NVIDIA_DRIVER_CAPABILITIES=all \
-v /tmp/.X11-unix:/tmp/.X11-unix \
mldarobotics/barn2024:april1
In the docker, the folder structure is similar to the suggested folder structure in the original README_BARN.md
# Navigate to this folder in the container
cd /jackal_ws/src/mlda-barn-2024/
# standard
python3 run.py --world_idx 0 --gui
# with rviz
python3 run_rviz.py --world_idx 0 --gui
All of 300 worlds .world
files have their initial camera angle changed for ease of viewing and troubleshooting
<gui fullscreen="0">
<camera name="user_camera">
<pose frame="">-2 4 10 0 1.57 3.14</pose>
<view_controller>orbit</view_controller>
<projection_type>perspective</projection_type>
</camera>
</gui>
DYNABarn is not used