• QGC/autopilot: vehicle drive and piloting
• Keep ROS and QGC responsibilities separate:
  • QGC bridge UDP/ffmpeg settings ros2 launch rov_onboard onboard_launch.py qgc_ip:=<CONTROL_LAPTOP_IP> pix_serial:=/dev/ttyACM0 pix_baud:=115200 onboard_launch.py now auto-starts MAVLink forwarding (mavlink-routerd) when QGC_IP is set or when qgc_ip:=... is passed directly: export QGC_IP=<CONTROL_LAPTOP_IP> If QGC does not show video: Run laptop: Run onboard:
• camera_node publishes /rov/camera/image_compressed on onboard.
• qgc_video_bridge_node on laptop subscribes and forwards low-latency H.264 over RTP to udp_host:udp_port (default port 5600) using ffmpeg.
• In QGroundControl, set video source to UDP and port 5600.
• If video does not decode in your QGC build, set output_format: h264 in src/rov_control/config/control_params.yaml as fallback.

5) Video to QGroundControl (via ROS bridge)

1. Connect QGroundControl to your autopilot (USB/telemetry Ethernet).
2. Calibrate joystick in QGroundControl.
3. Configure vehicle mode/buttons in QGroundControl.
4. Verify thruster response in QGC motor test / armed mode. Drive is handled outside of ROS by QGroundControl. To set up:

4) QGroundControl + Drive Setup (High Level)

• Also assumes you have QGroundControl installed on the control laptop.
  • qgc_video_bridge_node [QGroundControl] --MAVLink--> [Autopilot/ESCs] (drive control path, not ROS) ^ | UDP video :5600 from ROS bridge [rov_control/qgc_video_bridge_node] <-- ROS compressed image -- [rov_onboard/camera_node] Control Laptop (Ubuntu + QGroundControl + ROS 2) Onboard Computer (Ubuntu + ROS 2 + autopilot HW)

MATE 2026 ROV

UConn Underwater Robotics Club

MATEROV is an underwater robotics competition. This repository contains the ROS workspace for our 2026 entry, which includes:

• ROS 2 handles:
  • arm encoder values from the control laptop -> onboard servo commands
  • camera transport used by a QGC video bridge
• QGroundControl + autopilot stack handles vehicle driving/thrusters.
• QGroundControl + autopilot stack handles vehicle driving/thrusters.

---

Architecture

Control Laptop (Ubuntu + QGroundControl + ROS 2)          Onboard Computer (Ubuntu + ROS 2 + autopilot HW)

[QGroundControl] --MAVLink--> [Autopilot/ESCs]            (drive control path, not ROS)
      ^
      | UDP video :5600 from ROS bridge
[rov_control/qgc_video_bridge_node] <-- ROS compressed image -- [rov_onboard/camera_node]

[arm controller encoder USB] --/rov/arm/encoder_values--> [rov_control/arm_encoder_bridge_node]
                                                           |
                                                           '--/rov/arm/servo_command--> [rov_onboard/arm_servo_node -> PinPong servos]

---

ROS File Structure

Split into 2 main packages for the two hardware stacks, plus a shared messages package:

• rov_msgs
  • ArmServoCommand.msg
• rov_control (runs on control laptop)
  • arm_encoder_bridge_node
  • qgc_video_bridge_node
• rov_onboard (runs on onboard computer)
  • camera_node
  • arm_servo_node

---

ROS Topics (Current)

Topic	Type	Direction	Purpose
/rov/arm/encoder_values	std_msgs/Float32MultiArray	Laptop local input	8 encoder values from USB reader script/node
/rov/arm/servo_command	rov_msgs/ArmServoCommand	Laptop -> Onboard	8 servo target angles in degrees
/rov/camera/image_raw	sensor_msgs/Image	Onboard -> Laptop	raw camera stream
/rov/camera/image_compressed	sensor_msgs/CompressedImage	Onboard -> Laptop	compressed camera stream

ArmServoCommand axis order defaults to: [base, shoulder, elbow, wrist_pitch, wrist_roll, wrist_yaw, tool_rotate, gripper]

---

1) Install Dependencies

Run all commands on both machines unless noted.

• Assumes ROS 2 Humble and Ubuntu 22.04.
• Also assumes you have QGroundControl installed on the control laptop.

source /opt/ros/humble/setup.bash
sudo apt update
sudo apt install -y \
  python3-opencv \
  ros-humble-cv-bridge \
  python3-pip \
  ffmpeg \
  mavlink-router

On the onboard machine (for servos):

pip3 install pinpong

---

2) Build Workspace

Clone workspace from GitHub then run on both machines:

cd ~/MATE2026
source /opt/ros/humble/setup.bash
colcon build --symlink-install
source install/setup.bash

---

3) DDS Networking (ROS cross-machine)

Run on both machines (same value on both):

echo 'export ROS_DOMAIN_ID=42' >> ~/.bashrc
source ~/.bashrc

Quick check:

# machine A
ros2 topic pub /test std_msgs/String "data: hello"

# machine B
ros2 topic echo /test

---

4) QGroundControl + Drive Setup (High Level)

Drive is handled outside of ROS by QGroundControl. To set up:

1. Connect QGroundControl to your autopilot (USB/telemetry Ethernet).
2. Calibrate joystick in QGroundControl.
3. Configure vehicle mode/buttons in QGroundControl.
4. Verify thruster response in QGC motor test / armed mode.

---

5) Video to QGroundControl (via ROS bridge)

• camera_node publishes /rov/camera/image_compressed on onboard.
• qgc_video_bridge_node on laptop subscribes and forwards low-latency H.264 over RTP to udp_host:udp_port (default port 5600) using ffmpeg.
• In QGroundControl, set video source to UDP and port 5600.
• If video does not decode in your QGC build, set output_format: h264 in src/rov_control/config/control_params.yaml as fallback.

Run onboard:

cd ~/MATE2026
source install/setup.bash
ros2 launch rov_onboard camera_only_launch.py

Run laptop:

cd ~/MATE2026
source install/setup.bash
ros2 launch rov_control camera_viewer_only_launch.py

If QGC does not show video:

# verify compressed topic exists
ros2 topic list | grep /rov/camera/image_compressed

# verify stream rate
ros2 topic hz /rov/camera/image_compressed

---

6) Arm Control Path Setup

6.1 Encoder publisher contract (laptop)

There is a script/node that reads 8 encoder values from USB and publishes to ROS topic:

• topic: /rov/arm/encoder_values
• type: std_msgs/Float32MultiArray
• data length: 8

These values can be degrees or raw units. Mapping to servo angles is done in control_params.yaml (scales and offsets_deg).

6.2 Start arm bridge on laptop

cd ~/MATE2026
source install/setup.bash
ros2 launch rov_control arm_only_launch.py

6.3 Start onboard servo node

cd ~/MATE2026
source install/setup.bash
ros2 launch rov_onboard arm_only_launch.py

6.4 Enable real PinPong output

Edit src/rov_onboard/config/onboard_params.yaml:

• use_pinpong: true
• set pinpong_board and pinpong_port
• set servo_pins to your wiring
• tune min/max/neutral values per axis

Then rebuild onboard:

cd ~/MATE2026
source /opt/ros/humble/setup.bash
colcon build --symlink-install
source install/setup.bash

---

7) Full Launch (Simplified)

Onboard:

cd ~/MATE2026
source install/setup.bash
export QGC_IP=<CONTROL_LAPTOP_IP>
ros2 launch rov_onboard onboard_launch.py

onboard_launch.py now auto-starts MAVLink forwarding (mavlink-routerd) when QGC_IP is set or when qgc_ip:=... is passed directly:

ros2 launch rov_onboard onboard_launch.py qgc_ip:=<CONTROL_LAPTOP_IP> pix_serial:=/dev/ttyACM0 pix_baud:=115200

If needed, disable forwarding for bench tests:

ros2 launch rov_onboard onboard_launch.py enable_mavlink_forward:=false

Laptop:

cd ~/MATE2026
source install/setup.bash
ros2 launch rov_control control_launch.py

---

Key Config Files

• src/rov_control/config/control_params.yaml
  • encoder input topic
  • axis scaling/offset/clamps
  • QGC bridge UDP/ffmpeg settings
• src/rov_onboard/config/onboard_params.yaml
  • camera settings
  • PinPong + servo pin/range/timeout settings

---

Notes

• Keep ROS and QGC responsibilities separate:
  • ROS: arm + camera transport
  • QGC/autopilot: vehicle drive and piloting