Which drone to choose?
By Michal Barcis, Agata Gniewek, Roland Jung, and Petra Mazdin
The growing market for small drones has lead to the availability of new models for both consumer and commercial purposes. At the same time, there is an increasing interest of scientists to do drone research, which ranges from exploring the control of a single drone to establishing intelligent swarms of drones. The question is: Which drone to choose for our research activities in the context of our doctoral school on networked autonomous aerial vehicles? This question will be discussed in this short blog post.
When comparing different drone models, certain constraints need to be considered. Our work in multidrone systems puts a constraint in terms of the price per unit. Other relevant constraints and considerations are as follows: Release date, flight time length, computational power, take-off weight, payload (additional sensors or load), dimensions (a compromise between the payload and the size), low level controller specifications, wireless connections, sensors on board, open-source software, quality of documentation, possibility to customize a platform, and compatibility with PX4. These requirements are not the only ones but are the most relevant ones for our research.
The following short survey highlights our findings on four manufacturers and their products that we find worth examining:
Qualcomm Snapdragon
The Snapdragon is a computational unit for which a frame has to be purchased . It is fully featured with many sensors on board directly connected to a capable CPU. Documentation is currently limited. The architecture is relatively closed. The platform can be ordered together with the Erle Robotics drone mentioned below, where the Erle brain can be used instead of the Snapdragon. This requires assembling with the advantage of enabling us to construct a feature-rich copter.
Intel Aero Ready-To-Fly
The Intel Aero comes assembled. It offers open source support and detailed documentation. There are many sensors with four different cameras on board: 8 MP RGB front-facing, VGA down-facing, RealSense, and a monochromatic camera. It has an integrated, powerful companion board and the possibility to add an additional processor. We concluded that the Intel Aero drone offers a good value for its money (around 1000 €).
Erle Robotics
The Erle platform seems to be easy to customize. Raspberry Pi is used as a low level controller, which is very powerful. Cameras are accessible from the low-level controller. In conjunction with computation boards, such as Jetson TX2, the platform presents itself as a powerful candidate. The same processor that manages the control loop has access to all sensors and camera images. It might be that a low level controller running an OS might be more difficult to manage.
DJI M100
DJI is one of the leading manufacturers of small drones. The M100 comes ready-to-fly but might need additional sensors for our application. DJI offers a software development kit. One of the major drawbacks is the relatively high price.
