Aerial Robotics Landing Mission

Implementation of the final project for the EPFL Aerial Robotics course (MICRO-502, Prof. Dario Floreano).
The objective is to autonomously steer a Crazyflie nano‑drone from its take-off pad to a remote landing platform while mapping the environment, avoid obstacles, perform a precision landing, then take off again to return and land on the original platform.

Mission Highlights

• Structured flight pipeline – take-off, exploration, search, checkpoint landing, return flight, and final landing are handled via a state machine.
• Occupancy-grid mapping – builds a live grid from the four multiranger sensors and feeds the A* path planner.
• Local obstacle avoidance – combines reactive avoidance with path replanning and periodic yaw scans to refine the map.
• Search strategy – greedy coverage pattern over the exploration half of the arena to discover the landing pad.

Mission Demonstrations

Map generation walkthrough

Landing-pad acquisition

Repository Layout

hardware_if_research.py  # Main mission script
mission_utils.py         # Shared utilities (A*, distance helpers, geometry tools)
media/                   # Mission footage (final run, map walkthrough, landing search)

Requirements

Hardware:

• Crazyflie 2.x with Flow Deck v2 and Multi-Ranger Deck
• Crazyradio PA (or equivalent) for 2.4 GHz communication

Software:

• Python 3.9+
• cflib (Crazyflie radio interface)
• numpy

Optional:

• matplotlib if you enable map recording

Install dependencies with:

pip install cflib numpy matplotlib

Running the Mission

1. Connect the Crazyradio PA and ensure the Crazyflie is powered with the decks installed.

2. Adjust the URI in hardware_if_research.py if you are not using the default radio://0/20/2M/E7E7E7E7**.

3. (Recommended) Calibrate the Crazyflie sensors and verify the Flow deck height readings in the Crazyflie client.

4. Launch the mission:

   python hardware_if_research.py

5. The script will:

   • Reset the Kalman filter,
   • Ascend to the target altitude,
   • Navigate to the exploration wall,
   • Search and land on the intermediate pad,
   • Take off again and return to the start.

6. Upon completion the motors are disarmed and the drone remains on the final platform.

If you want to observe the generated occupancy grid, set map_recording=True in MissionConfig. The current implementation keeps the hook ready but does not write files by default.

Media

• Mission footage lives in media/.
• media/final_run_vid.mp4 – A short clip of the real run covering map generation, landing-pad detection, and the return flight.

Extending the Project

• Plug a SLAM back-end or external motion capture data to improve localisation.
• Log and visualise the occupancy map over time.
• Replace the greedy search with frontier-based exploration to speed up pad discovery.
• Add safety checks (battery, timeout) before initiating the return flight.

Acknowledgements

This work was produced as part of the Aerial Robotics class at EPFL.
Many thanks to the course staff and teammates for feedback, testing time, and arena setup support.