Software Setup v2

Intro

Update to Software Setup v1

Setup

Install Dependencies

• Install pip dependencies

pip3 install "ultralytics==8.3.176" "numpy<2.0" pandas --break-system-packages 

• Install apt dependencies ROS2 Jazzy

sudo apt install ros-jazzy-robot-localization ros-jazzy-tf-transformations

Pontus Installation

• Install Pontus

cd mrg_ws/src
git clone https://github.com/gt-marine-robotics-group/Pontus.git

Simulation

• Launch odom simulation

ros2 launch pontus_bringup odom_simulation.launch.py 

Example of Gazebo configured for Sonar Data Visualization

Sensor Visualization & Debugging

Gazebo

• Visualize simulated sonar hit locations

RViz

• Run RViz

rviz2

Perception

• Launch Perception nodes

ros2 launch pontus_perception perception.launch.py 

ros2 launch pontus_perception detection_services.launch.py auv:=sim

Mapping/Localization

Uses perception outputs + robot state (DVL + EKF) to maintain a global estimate of the environment.

ros2 launch pontus_mapping mapping.launch.py 

Controls

Command Mode

The command mode is set by publishing to the command_mode topic

E-Stop

• Mode 0

Direct Control

• Mode 1- Control vehicle by publishing to cmd_vel

Velocity Controller

• Mode 2: Matches cmd_vel

Position Controller

• Mode 6: Position Face Travel- Listens on cmd_pos to move to correct depth, orient towards target, move to target, then turn to match commanded orientation
• Mode 7: Position with Strafe- Move to directly to commanded pose on the cmd_pos topic

RQT

• Run rqt (To display topics: Click plugins → topics → message publisher)

rqt

Position Controller

• In RQT, select command_mode and click the + icon to add the topic
• Set the mode by modifying the expression field to 6 or 7
• Add cmd_pos to test the position controller (Ex: set pos x to 2)

Plotjuggler

• Install plotjuggler

sudo apt install ros-humble-plotjuggler-ros

• Run plotjuggler

ros2 run plotjuggler plotjuggler

• When the window loads, click start, and select topics, then drag variables (e.g. /cmd/linear/x) onto the main plot area

• Testing position controller

• depth controller: linear z

  • going against buoyancy
  • looks at position moving away from where it should be and makes adjustment to push it down

• not running SLAM

  • Get position from DVL and robot_localization runs extended kalman filter

Autonomy

• Run autonomy nodes