Adding autonomous capabilities to a inexpensive quadcopter by:
- Reverse engineering the communications between the drone and it's controller to control it from a personal computer.
- Learning the rigid-body dynamics of the quadcopter.
- Learning to estimate motion and depth from monocular image sequences.
- Path planning in the depth image space and controls using the learned dynamics model and MPPI.
Drone model: F180D
This drone model may be discontinued, though some can be found for second-hand sale online. The methods used to gain control of this drone can likely be reproduced on other inexpensive hobby drones.
Packet Sniffer
Decoding packets: Controls, Camera
Data collection setup, model
UnDeepVO