Repository for the ROS software that runs on the AAU modified ExoMy Rover.
This version of the system runs on a NVIDIA Xavier, using Ubuntu 20.04, jetpack 5.0 and ROS Foxy.
To launch the camera nodes, open 2 terminals and run each of these lines:
ros2 launch realsense2_camera rs_launch.py camera_name:=cam_1 serial_no:="'912112070533'" device_type:=d435 depth_module.profile:=424x240x30 hole_filling_filter:=true
ros2 launch realsense2_camera rs_launch.py camera_name:=cam_2 serial_no:="''" device_type:=d435 depth_module.profile:=424x240x30 hole_filling_filter:=true
Documentation about the RealSense ROS nodes can be found at their github
Lastly, run the system in a third terminal:
ros2 launch exomy exomy.launch.py
This ROS system depends on certain ROS2 packages, we have not been able to install compiled versions on the ARM64 processer, therefore we have compiled the packages ourselves.
ros2_numpy
realsense-ros
sensor_msgs
rviz2 (for debugging)
These packages where cloned from their respected Github repositories and built using colcon.
Python dependencies:
scipy
numpy
math
time
torch
csv
adafruit_pca9685
gym
skrl
The ROS system has 4 main nodes:
Camera transformation node
RL node
Kinematics node
Motor node
The nodes are located in ~/ExoMy_Software/exomy
The nodes are written in python, and have supportive scripts located in ~/ExoMy_Software/scripts/utils
Further is nodes called joy and gamepad_parser, these are for controlling the rover with a controller, but is disabled at default.
If you wish to include these, edit the launch file located at ~/ExoMy_Software/launch/exomy.launch.py
When making changes to either the nodes or subscripts, the ROS system has to be recompiled, do this by opening a terminal in ~/ExoMy_Software and running:
colcon build
If you use our Xavier, we have used virtual python environments, so if you have problems compiling or other python problems you can try switching between then with:
pyenv local system
pyenv local 3.8.0
These commands switch between the two environments on the system.
For controlling the robot wirelessly, we have used SSH to control and program the rover. The Eduroam AAU network does not allow communication directly between devices however. We have used a free VPN called tailscale. We strongly recommend using this, as you can share the device, such that all group members can access it.
For other questions, contact
- Anton: [email protected]
- Emil: [email protected]
- Jacob: [email protected]