docker-sgp-tools

docker-sgp-tools provides ready-to-use Docker Compose files and Dockerfiles for running SGP-Tools and its ROS2 companion package in two main scenarios:

• Simulation: Full-stack GUI container (Gazebo, ArduPilot SITL, ROS2, Foxglove, SGP-Tools).
• Robot/Edge/BlueOS: Minimal ROS2/SGP-Tools container, usable directly on a robot or as a BlueOS extension.

Containers & Compose Files

Compose File	Use Case	Features
sim-compose.yml	Simulation/Dev	GUI, ROS2, Gazebo, ArduPilot SITL, SGP-Tools, Foxglove. Requires Nvidia GPU.
robot-compose.yml	Robot/Production	Minimal ROS2/SGP-Tools. Runs on robot or deployable as a BlueOS extension.

Prerequisites

• Docker Engine
• Docker Compose
• NVIDIA Container Toolkit (for GPU support in simulation)
• Optional for multi-arch builds: Buildx

Getting Started

1. Clone this repository

git clone --recurse-submodules https://github.com/itskalvik/docker-sgp-tools.git
cd docker-sgp-tools

2. Pull and Run the Simulation Container

docker compose -f sim-compose.yml pull
docker compose -f sim-compose.yml up -d
docker compose -f sim-compose.yml exec sgptools bash

To stop or clean up:

docker compose -f sim-compose.yml stop    # stop containers, keep data/volumes
docker compose -f sim-compose.yml down    # stop and remove containers/volumes/networks/images

3. Run the Minimal Robot Container

For robot deployment onboard a custom ASV:

docker compose -f robot-compose.yml up -d
docker compose -f robot-compose.yml exec sgptools bash

• The robot-compose.yml file can be customized for device paths, data log locations, and environment variables as needed.

• See the BlueOS SGP-Tools README for step-by-step instructions on deploying as a BlueOS extension on a Blue Robotics Blue Boat.

Simulation Workflow Example

1. Launch Gazebo simulation with BlueBoat ASV:

gz sim -v4 -r blueboat_waves.sdf

2. Start ArduPilot SITL (inside container):

sim_vehicle.py -v Rover -f rover-skid --model JSON --console --map -N -L RATBeach

• Wait for AHRS and GPS to turn green in MAVProxy.

• If errors (paramftp: bad count), restart sim_vehicle.py.

3. Launch SGP-Tools planner:

ros2 launch ros_sgp_tools asv.launch.py

4. (Optional) Visualize in Foxglove:

• Open Foxglove web (https://app.foxglove.dev/)
• Connect to ROS2 bridge at ws://localhost:8765

About the Images

• Simulation container:

  • Ubuntu, ROS2 Humble, Gazebo, MAVProxy, ArduPilot SITL, Foxglove bridge, SGP-Tools, all required plugins and environment scripts.

  • Includes ros_entrypoint_sitl.sh to automatically set up Gazebo/ArduPilot/ROS2 plugin paths on container start.

  • NVIDIA GPU required for Gazebo reliability.

• Robot container:

  • Lightweight image: ROS2, SGP-Tools, minimal ROS2 nodes, sensor drivers.

  • Designed for ARM64 and AMD64 targets (supports BlueOS and bare-metal robot computers).

Building from Scratch (Multi-Arch)

To build containers for both ARM64 and AMD64:

docker buildx create --name multi-arch --platform "linux/arm64,linux/amd64" --driver "docker-container"
docker buildx use multi-arch

docker compose -f sim-compose.yml build
# or
docker compose -f robot-compose.yml build

Robot/BlueOS Extension

• The robot container can be deployed as a BlueOS extension on compatible vehicles (e.g., BlueRobotics BlueBoat).

• See the BlueOS SGP-Tools README for detailed instructions, swap size configuration, file upload, and mission launch procedures.

References

• Based on Dockerfiles from:
  • https://github.com/Robotic-Decision-Making-Lab/blue
  • https://github.com/ryomo/ros2-gazebo-docker