A complete simulation environment for UAV (drone) and UGV (ground vehicle) development using GPS-denied navigation with vision-based localization, while maintaining GPS-based geofencing for safety.

Navigation Mode: All vehicles navigate using relative positioning only:

✅ Visual odometry from cameras
✅ Optical flow sensors
✅ IMU integration
✅ Visual landmark tracking
✅ Local coordinate frames

GPS Usage: GPS is ONLY used for:

✅ Geofencing (safety boundaries)
✅ Initial position reference (optional)
❌ NOT used for waypoint navigation
❌ NOT used for position control

This simulates real-world GPS-denied environments like:

Indoor spaces
Urban canyons
GPS-jammed areas
Under bridges/tunnels

Features

🚁 Iris quadcopter with dual cameras (forward + downward)
🚗 Ground vehicle (UGV) with vision sensors
📷 Visual odometry - camera-based position estimation
👁️ Optical flow - velocity estimation from downward camera
🗺️ Landmark navigation - visual feature tracking
🛡️ GPS geofencing - safety boundaries only
🎮 Single command launch - everything runs from one script
🖥️ NVIDIA GPU acceleration support
🐍 Python virtual environment for isolated dependencies
🌍 GPS-denied worlds - indoor and urban environments

System Requirements

Ubuntu 22.04 LTS
Python 3.10
ROS 2 Humble
8GB RAM minimum (16GB recommended)
NVIDIA GPU recommended

Quick Start

# 1. Clone repository
cd ~/ros2_ws/src
git clone <your-repo-url> uav_ugv_simulation
cd uav_ugv_simulation

# 2. Run setup (installs everything)
bash setup.sh

# 3. Restart terminal or reload bash
source ~/.bashrc

# 4. Activate virtual environment and run
source activate_venv.sh
bash scripts/run_simulation.sh

┌─────────────────────────────────────────────────────┐
│              Vision Sensors                         │
│  Forward Camera + Downward Camera + Optical Flow    │
└─────────────────┬───────────────────────────────────┘
                  │
                  ▼
┌─────────────────────────────────────────────────────┐
│         Visual Odometry & Feature Tracking          │
│  Estimates relative position from camera motion     │
└─────────────────┬───────────────────────────────────┘
                  │
                  ▼
┌─────────────────────────────────────────────────────┐
│         Position Estimator (EKF Fusion)             │
│  Fuses: Visual Odom + Optical Flow + IMU            │
│  Output: Local position estimate (relative)         │
└─────────────────┬───────────────────────────────────┘
                  │
                  ▼
┌─────────────────────────────────────────────────────┐
│            Navigation Controller                     │
│  Commands: "Move 5m forward, 3m right"              │
│  (Relative coordinates only, NO GPS waypoints)      │
└─────────────────────────────────────────────────────┘

         SEPARATE SAFETY LAYER (GPS-based):
┌─────────────────────────────────────────────────────┐
│              Geofence Monitor                        │
│  GPS position → Check against boundaries            │
│  If outside: Emergency RTL or hold                  │
└─────────────────────────────────────────────────────┘

1. Vision-Only Mode (Default)

Uses camera for all position estimates
Suitable for structured environments
Requires good lighting and visual features

2. Optical Flow Mode

Uses downward camera for velocity
Works well at low altitudes
Good for hovering and slow flight

3. Hybrid Mode

Combines visual odometry + optical flow + IMU
Most robust approach
Recommended for complex missions

Geofencing Configuration

Edit config/geofence_params.yaml:

geofence:
  enabled: true
  use_gps: true  # GPS ONLY for geofence
  
  # Define boundaries (GPS coordinates)
  fence_type: "polygon"  # or "circle"
  
  # Polygon fence (lat/lon points)
  polygon_points:
    - {lat: 47.397742, lon: 8.545594}  # Point 1
    - {lat: 47.398242, lon: 8.545594}  # Point 2
    - {lat: 47.398242, lon: 8.546094}  # Point 3
    - {lat: 47.397742, lon: 8.546094}  # Point 4
  
  # Or circle fence
  center_lat: 47.397742
  center_lon: 8.545594
  radius_meters: 100
  
  # Actions on breach
  action: "RTL"  # Return to launch
  max_altitude: 50  # meters

Example Mission (Relative Coordinates)

# Example: Navigate to visual landmark

# Define mission in RELATIVE coordinates
mission_waypoints = [
    {"x": 0, "y": 0, "z": 5},      # Takeoff to 5m
    {"x": 10, "y": 0, "z": 5},     # Move 10m forward
    {"x": 10, "y": 5, "z": 5},     # Move 5m right
    {"x": 0, "y": 5, "z": 5},      # Return to start (offset)
    {"x": 0, "y": 0, "z": 5},      # Back to takeoff point
    {"x": 0, "y": 0, "z": 0},      # Land
]

# GPS is NEVER used for these waypoints
# Position estimated from visual odometry

Project Structure

launch/ - ROS 2 launch files
worlds/ - Gazebo world files (indoor, urban)
models/ - Robot models (Iris with cameras, UGV)
src/vision/ - Visual odometry, optical flow
src/localization/ - Position estimation, sensor fusion
src/navigation/ - Path planning (relative coordinates)
src/safety/ - Geofencing (GPS-based)
config/ - Configuration files

Documentation

Aspect	GPS Navigation	This Project (GPS-Denied)
Position Source	GPS satellites	Visual odometry + sensors
Waypoint Type	GPS coordinates	Relative coordinates (x,y,z)
Reference Frame	Global (lat/lon)	Local (relative to start)
Indoor Capability	❌ No	✅ Yes
Drift	Minimal	Accumulates over time
Geofencing	GPS-based	GPS-based (safety only)
Use Cases	Outdoor, open sky	Indoor, urban, GPS-jammed

Running Different Scenarios

# Indoor warehouse (no GPS available)
bash scripts/run_simulation.sh --world worlds/indoor_warehouse.world

# Urban canyon (degraded GPS)
bash scripts/run_simulation.sh --world worlds/urban_canyon.world

# Open outdoor (GPS available but not used for nav)
bash scripts/run_simulation.sh --world worlds/empty_custom.world

License

MIT License

Contributing

Contributions welcome! Please ensure all navigation remains GPS-denied (except geofencing).

README.md

UAV-UGV Gazebo SITL Simulation

GPS-Denied Navigation with Geofencing

🎯 Key Feature: GPS-Denied Navigation