Skip to the content.

Back

Beacon-Based 3D Localisation of a Quadcopter Using Extended Kalman Filter

We estimated the 3D position of a quadcopter with an Extended Kalman Filter, fusing a motion model with distance measurements from static RF beacons. It runs in Python on ROS Noetic, simulated in Webots, with the filter itself built on FilterPy.

The beacons are Webots emitter and receiver pairs, each on its own channel so that the drone can tell them apart. Distance is estimated from signal strength, which is noisy, and that noise is the reason the filter is there in the first place. We ran it with both the Crazyflie and the DJI Mavic 2 PRO models.

The estimate held up in simulation under sensor noise, changes to the beacon layout and loss of signal.

The code is on GitHub, and the report is here, written in Spanish.

simulation

Tech Explored

Highlights

Lowlights

Lessons Learned

Cool Drones

Here are some pictures of cool drones that were in the Griffin project laboratory:

Big Boy

Tunnel Inspection Boy

Birdie

Blurry Pipe Boy