This repository is published and maintained by the Volkswagen Group Machine Learning Research Lab.
Learn more at https://argmax.ai.
This repository contains the source code of mechanical parts used for the construction of a quadrotor frame suitable for machine learning and control research. The requirements for such a frame are:
- Protect propellers and internal parts from crashes and drops from midair
- Possibility to mount sensors around the perimeter
- Easy to manufacture in-house
- Lightweight structure
This resulted in the following frame design.
All models for digital manufacturing can be compiled using the included makefile. The program openscad needs to be installed on the system. The files are stored in the build folder after running the following command.
make -j
The following parts should be printed in PLA on a 3D printer.
1x Holder for Battery | 4x Mount for Motors |
---|---|
![]() |
![]() |
24x Holder for Distance Sensors | 1x Mount for Raspberry Pi |
---|---|
![]() |
![]() |
The following parts should be cut out of 2 mm thick polyamide with a laser cutter.
1x Drone Cage Top | 1x Drone Cage Bottom |
---|---|
![]() |
![]() |
4x Drone Cage Sides | 4x Drone Cage Pillars |
---|---|
![]() |
![]() |
If you find the code or models in this repository useful for your research, please consider citing our work. The quadrotor frame described in this repository is used in the following paper and dissertation.
@misc{beckerehmck2020learning,
title={Learning to Fly via Deep Model-Based Reinforcement Learning},
author={Philip Becker-Ehmck and Maximilian Karl and Jan Peters and Patrick van der Smagt},
year={2020},
eprint={2003.08876},
archivePrefix={arXiv},
primaryClass={cs.RO}
}
@phdthesis {karl2020unsupervisedcontrol,
author={Maximilian Karl},
title={Unsupervised Control},
type={Dissertation},
school={Technische Universität München},
address={München},
year={2020}
}
The purpose of this source code is limited to bare demonstration of the experimental section of the related papers.