naterpotatoers/Splot

Autonomous robot for AMD's Pervasive AI Contest

Splot

A quadruped robot built for AMD's Pervasive AI Contest 2024.

Installing

• Compute platform: Raspberry Pi 4 8Gb Model
• Operating system: Ubuntu Server 22.04.3 LTS (64-bit)
• Python: 3.10.12
• ROS: ROS2 Iron Irwini (follow the installation instructions and choose ros-iron-ros-base)

Run git clone --recurse-submodules [git repo url] to clone this repository with submodules (i.e. joystick_drivers).

For gamepad support, follow the installation instructions on the joystick_drivers wiki. Additionally, apt install ros-iron-diagnostic-updater libx11-dev libxext-dev.

Install python dependencies using python3 -m pip install -r Splot/requirements.txt. Note that legacy code using the PCA9685 boards also requires adafruit-circuitpython-servokit.

Building

• Run sudo rosdep init to initialize rosdep if you have not done so already.
• Run rosdep update to update rosdep index
• For each of Splot/ros2/ros2_joy_ws, Splot/ros2/ros2_robot_localization_ws, and Splot/ros2/ros2_ws:
• cd into the workspace directory (i.e. Splot/ros2/ros2_robot_localization_ws)
• Run rosdep install --from-paths src -y --ignore-src to install dependencies
• Run colcon build --symlink-install to build the packages

Running

• Source the ROS workspaces using source Splot/ros2/ros2_ws/install/local_setup.bash, source Splot/ros2/ros2_robot_localization_ws/install/local_setup.bash, and source Splot/ros2/ros2_joy_ws/install/local_setup.bash (these need to be run every time you login/reboot)
• cd into Splot/ros2/launch and run ros2 launch mission_control_joy.yaml to run the gamepad input demo, or choose a different launch file

Dependencies

• joystick_drivers - ROS2 nodes for receiving joystick/gamepad input
• robot_localization - ROS2 sensor fusion package implementing extended kalman filtering and unscented kalman filtering
• adafruit-circuitpython-bno08x - Python package for interfacing with BNO08x family 9-axis absolute orientation IMUs
• adafruit circuitpython-gps - Python package for interfacing with Adafruit's PA1616D "Ultimate GPS Breakout"
• pyserial - Python package for serial communication

Other Notes

By default, Ubuntu does not allow access to the Raspberry Pi's built-in RX/TX serial pins, which we use to communicate with the GPS. We followed these instructions to enable hardware serial access.

Scout

Scout uses MAVLINK to communicate with the drone. The MAVLINK code is written in C++ and is attached to the python script that is used for the search algorithm. The C++ code is located in the MAVLINK/examples folder, but first needs to be copied and pasted into the MAVLINK/examples folder. Then run the python script in scout/demo_missions/scout_search.py and the C++ code will be executed. This script is expected to be connected to a simulation such as jmavsim, gazebo, or a real drone. (IP must be changed in the C++ script to connect to the real drone)

Dependencies

update all submodules by running git submodule update --init --recursive

PX4 Simulation

To run the PX4 simulation, follow the instructions in the PX4 User Guide. The px4 package is included as a submodule in this repository.

TODO List:

• Add more python functionalities for more drone control and autonomy.
• Add IP change in python so we can easily connect to the real drone or the simulation.
• Add offboard autonomous control when needed.
• Add Camera functionality on drone.
• Add CNN for object/person detection.
• Add autonomous search and rescue algorithm.