7.6 KiB
7.6 KiB
Docker Setup for RDC Simulation
This guide explains how to run the RDC Simulation environment inside a Docker container with NVIDIA GPU support and X11 display forwarding. This is the recommended approach for Ubuntu 25.x+ or any system where native installation has dependency issues.
Benefits of Docker
| Aspect | Native Install | Docker Container |
|---|---|---|
| Ubuntu 25.x support | ❌ Dependency conflicts | ✅ Works perfectly |
| ROS 2 Jazzy | ❌ No packages for 25.x | ✅ Full support |
| Gazebo dev packages | ❌ Library conflicts | ✅ All packages work |
| Host system | Modified (packages, bashrc) | ✅ Untouched |
| Reproducibility | Varies by system | ✅ Identical everywhere |
Prerequisites
1. Docker Engine
# Install Docker
curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER
newgrp docker # or log out and back in
2. NVIDIA GPU & Drivers (Required for Gazebo)
# Check your GPU is detected
nvidia-smi
3. NVIDIA Container Toolkit
# Install NVIDIA Container Toolkit
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker
Verify GPU is accessible in Docker:
docker run --rm --gpus all nvidia/cuda:12.0-base nvidia-smi
Quick Start
1. Build the Image
cd ~/RDC_Simulation
docker compose build
Note: First build takes 20-40 minutes (downloads ~5GB, compiles ArduPilot + plugins).
2. Allow X11 Display Access
xhost +local:docker
Add this to your
~/.bashrcto make it permanent.
3. Run the Container
docker compose run --rm simulation
This drops you into a bash shell inside the container as user pilot.
Running the Simulation
Once inside the container, the environment is pre-configured. You'll see a welcome message with available commands.
Option 1: All-in-One Script
./scripts/run_ardupilot_controller.sh
Option 2: Multi-Terminal (Recommended for Development)
Terminal 1 - Start Gazebo:
./scripts/run_ardupilot_sim.sh runway
Terminal 2 - Open new shell in same container:
# On host, find container and exec into it
docker exec -it rdc-sim bash
# Then start SITL
sim_vehicle.py -v ArduCopter -f gazebo-iris --model JSON --console
Terminal 3 - Open another shell:
docker exec -it rdc-sim bash
# Run controller
python scripts/run_ardupilot.py --pattern square
Docker Commands Reference
Build Commands
# Build image (standard)
docker compose build
# Build with no cache (clean rebuild)
docker compose build --no-cache
# Build just the Dockerfile directly
docker build -t rdc-simulation .
Run Commands
# Interactive shell
docker compose run --rm simulation
# Run specific command
docker compose run --rm simulation ./scripts/run_ardupilot_sim.sh runway
# Run headless (no display, for CI)
docker compose run --rm simulation-headless
# Start container in background
docker compose up -d simulation
docker exec -it rdc-sim bash
Cleanup Commands
# Stop all running containers
docker compose down
# Stop and remove containers
docker compose down --remove-orphans
# Remove built image (to rebuild from scratch)
docker rmi rdc-simulation:latest
# Remove all project containers and images
docker compose down --rmi all
# Full cleanup (removes image + build cache)
docker compose down --rmi all
docker builder prune -f
Complete Uninstall & Rebuild
To completely remove everything and start fresh:
# 1. Stop any running containers
docker compose down
# 2. Remove the image
docker rmi rdc-simulation:latest
# 3. Clear Docker build cache (optional, saves disk space)
docker builder prune -f
# 4. Clear unused Docker data (optional, aggressive cleanup)
docker system prune -f
# 5. Rebuild from scratch
docker compose build --no-cache
Video Recording
The Docker image includes ffmpeg and tools to record the simulation.
Record Flight
python scripts/record_flight.py --pattern square --quality high --output my_flight
Videos are saved to recordings/ inside the container.
Copy Recordings to Host
# From host machine
docker cp rdc-sim:/home/pilot/RDC_Simulation/recordings .
Headless Mode (CI/Server)
For running on servers without a display:
docker compose run --rm simulation-headless
Gazebo can still render offscreen for video recording using software rendering.
Configuration Reference
docker-compose.yml Services
| Service | Use Case |
|---|---|
simulation |
Full GUI with X11 display |
simulation-headless |
No display, for CI/testing |
Environment Variables
| Variable | Purpose |
|---|---|
DISPLAY |
X11 display (auto-detected) |
NVIDIA_VISIBLE_DEVICES |
GPU visibility |
HEADLESS |
Set to 1 for software rendering |
Volumes
| Mount | Purpose |
|---|---|
/tmp/.X11-unix |
X11 socket for display |
~/.Xauthority |
X11 authentication |
Troubleshooting
"permission denied" on docker commands
sudo usermod -aG docker $USER
newgrp docker # or log out and back in
"Could not select device driver 'nvidia'"
NVIDIA Container Toolkit not installed or configured:
# Reinstall toolkit
sudo apt-get install -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker
"All CUDA-capable devices are busy"
# Check GPU on host
nvidia-smi
# Make sure no other process is using all GPU memory
"Could not connect to display :0"
# On host machine, allow X11 connections
xhost +local:docker
# If using SSH, ensure X11 forwarding is enabled
ssh -X user@host
"Error response from daemon: No such image"
Build the image first:
docker compose build
Build fails at ArduPilot step
The ArduPilot build requires significant memory. If it fails:
# Increase Docker memory limit (Docker Desktop)
# Or reduce parallel jobs in Dockerfile:
# Change: make -j$(nproc)
# To: make -j2
Container runs but Gazebo crashes
# Check OpenGL inside container
docker compose run --rm simulation glxinfo | grep "OpenGL renderer"
# Should show your NVIDIA GPU, not "llvmpipe" (software)
Development Workflow
Edit Code on Host
Your code changes in ~/RDC_Simulation are not automatically reflected in the container (unless you mount the volume). To test changes:
# Option 1: Rebuild image (slow)
docker compose build
# Option 2: Mount your code (edit docker-compose.yml to add volume)
volumes:
- .:/home/pilot/RDC_Simulation
Build Custom Image Tag
docker build -t rdc-simulation:dev .
docker run --gpus all -it rdc-simulation:dev
What's Installed in the Container
| Component | Version |
|---|---|
| Ubuntu | 24.04 LTS |
| ROS 2 | Jazzy |
| Gazebo | Harmonic |
| Python | 3.12 |
| ArduPilot SITL | Latest (git) |
| ArduPilot Gazebo Plugin | Latest (git) |
| MAVProxy | Latest (pip) |
All Gazebo development packages are included for plugin compilation.