Please visit our website and paper for more information about this project.
This repo contains instructions for how to use the Isaac SIM simulation for the Atron USV autonomous debris collection vehicle.
mkdir -p ros2_ws/src
cd src
git clone https://github.com/RISC-NYUAD/ATRON
cd ../
rosdep install --from-path src --ignore-src -r -y
colcon build --symlink-install && source install/setup.bash
echo "source ~/ros2_ws/install/setup.bash" >> ~/.bashrc
- Run the simulation in Isaac Sim using the file in
isaac/atron_simulation.usd
- Run the autonomous routine using:
ros2 launch triton_bringup simulation.launch.xml
- In another terminal, run
ros2 launch triton_navigation op_solver_bridge.launch.py
This will launch the planning algorithms, as create a link to the orienteering solver.
- Run the following in another terminal to run the Orienteering -> OMPL -> Pure pursuit navigation pipeline.
cd ~/ros2_ws/src/Triton/triton_navigation/config
ros2 service call /export_oplib std_srvs/srv/Trigger
./../../dependencies/op_solver_ros/op-solver/build/src/op-solver opt --op-exact 0 problem.oplib
ros2 service call /create_waypoints std_srvs/srv/Trigger
ros2 service call /generate_path std_srvs/srv/Trigger
ros2 service call /navigate std_srvs/srv/Empty
We also provide a high-fidelity simulation which includes improved visual and physical details.
Buoyancy is simulated via force calculations on a height wavefunction. Movemement is implemented via thrust forces controlled via PID to execute the desired velocity.
See the releases tab for the atron_highfidelity for the simulation file. The simulation can physically simulate sea states from 0 to 4 on the Beaufort Scale
We also implement a correction step for the debris and obstacle localization process. Below are videos showing the before and after for the correction step.
Without the correction in a buoyancy environment (cylinders are for the obstacle class, spheres for the debris class):
With the correction:
The photorealistic environment also allows the training of object detection models, as seen below
