Project ODYSSEY is an advanced vision-based navigation framework designed for UAV operations in GPS-denied environments (indoors, forests, urban canyons). It leverages Visual-Inertial Odometry (VIO) and 3D SLAM to enable fully autonomous flight where traditional satellite navigation fails.
Standard drones rely on GPS to know their position. Project ODYSSEY solves the "Lost Drone" problem by using a Stereo Depth Camera and IMU Fusion to build a map of the environment and calculate its own position relative to obstacles in real-time.
- Middleware: ROS 2 Humble / Jazzy
- Simulation: Gazebo Harmonic (Ignition)
- Visual SLAM: ORB-SLAM3 / RTAB-Map
- Perception: OpenCV / Point Cloud Library (PCL)
- Path Planning: Nav2 (Navigation 2 Stack)
src/: Custom ROS2 nodes for sensor fusion and vision processing.models/: URDF/SDF descriptions of the quadcopter with stereo-vision sensors.worlds/: Complex Gazebo environments (Forest, Industrial Warehouse).launch/: Python launch files for the complete autonomy stack.
- Phase 1: Integrated ROS2-Gazebo sensor bridge with 100% telemetry sync.
- Phase 2: Real-time 3D Occupancy Grid generation from depth data.
- Phase 3: < 0.5m drift in VIO-only flight missions.
- Phase 4: Fully autonomous dynamic obstacle avoidance.
Developed by Yogesh E S
*Aerospace Portfolio -