Cyclone Hybrid Vehicle (Tailsitter)

Together, the Dronelab at ENAC in Toulouse and the MAVLab at TU Delft have developed the Cyclone. It is a hybrid drone of the tailsitter type, with just 4 actuators. In the video, we show autonomous flight, ranging from hover to forward flight. Below, we will highlight some of the interesting aspects, but more details can be found in the paper.

The vehicle is designed with efficiency and simplicity in mind. More than anything else, the Cyclone is a fixed wing aircraft capable of hovering. It is equipped with two propellers and two flaps. To spare the beautiful plane in the image below, we made a cheap version to test with. The Cyclone that is starring in the video has a 3D printed fuselage and a foam wing laminated with Kevlar. With the battery we are currently flying with it weighs 1.1 kg, flies most efficient at 16 m/s, and tops out at about 30 m/s (level flight).

To control the Cyclone, we use a method called Incremental Nonlinear Dynamic Inversion (INDI). The benefit of this method is that it is able to compensate for unmodeled effects or disturbances, by considering the angular acceleration. After all, according to Newton, all moments acting on the drone together result in an angular acceleration. This way, it is possible to compensate these moments and disturbances without modeling them.


The way we see it, a hybrid, like any rotorcraft needs to be able to keep its position regardless of wind. This means that the drone needs to be able to fly at any constant airspeed (0-30 m/s) without ascending. Therefore, were for a transition to forward flight or back, climbing a bit will make things much easier, we are not doing this. However, that means that we need to deal with high angle of attack flight, which generates a large pitch-down moment. A moment so large, that saturation of the flaps is commonplace (watch the flaps in the video during the back-transition).

We used the newly designed 'Chimera' autopilot, which is aimed at ease of use and runs the Paparazzi open source autopilot software. Among the many features are an STMF7 processor, an SD card slot, a mount for an XBEE and a differential pressure sensor.

Views: 569

Comment by Hector Garcia de Marina on July 19, 2017 at 2:37am

The sideways transition is my favorite in this vehicle, so smooth :P.

Comment by Vladimir "Lazy" Khudyakov on July 19, 2017 at 7:45am

Nice project! 

Comment by Bart Theys on July 27, 2017 at 5:52am

Very nice Ewout! So your control is (much?) better than the control of the Wingtra?

Comment by Ewoud Smeur on July 27, 2017 at 7:01am

Thanks Bart! The Wingtra is also a cool project, but I think they don't publish anymore since they started a company. That is why I don't think there is enough information available for such a conclusion, based solely on this paper: Here they describe what is in essence a PD controller.

I think the INDI approach offers advantages in compensating aerodynamic forces and moments that are difficult to model, and in rejecting disturbances. In the videoclip, you can see the very slow transition, showing controlled flight at every speed, without gaining altitude. Traversing the different flight regimes, the pitching moment changes significantly. Without modelling this, it is handled by the INDI controller. In the Wingtra videos I only see quick transitions, so either hover or forward flight.

Comment by Hector Garcia de Marina on July 27, 2017 at 7:22am

Hi Bart,

The attitude of the tailsitter Wingtra is controlled by a "geometric controller" . No idea whether they are still employing the same algorithm after the paper that Ewoud cited. This controller has been proposed first for satellites? and later on implemented in rotors .  It is a PD controller, but the error signals are not just the attitude angles or alike, the "trick" for generating such signals is quite mathematical profound. It has very nice properties, you can calculate regions of attraction for the stability (quite big, you can flip around your vehicle) and guarantee exponential convergence. On the other hand, the controller requires to have some knowledge (models) about the momentum generated by your actuators.  

Ewoud will confirm it. Very roughly speaking the INDI does not require a model of the actuators, just it is asking for "more or less" momentum/thrust (like an integrator) but not precisely for a number (like in the geometric since it is based on a P controller but with a fancy error signal) in order to achieve the desired accelerations (the accelerations are your error signals in the INDI).


You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2017   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service