Last academic year at KU Leuven, we designed, built and test flown a VTOL UAV: the VertiKul. During this project we gratefully made use of the info and support of the DIY Drones community and therefore we would like to share our results on this project.
The VertiKul is designed for automated aerial transport of small packages and is optimized for maximum range and payload capability. The innovative design makes use of the benefits of both multi-rotors and fixed-wing airplanes. For take-off and landing, the VertiKul hovers like a quadrotor and for forward flight, the VertiKul pitches 90° and flies like an airplane.In airplane mode, the attitude is also controlled by differential thrust of the motors. Therefore, no additional control surfaces are required, reducing the number of moving parts, risk of failure and maintenance cost. The structure is made out of three carbon fiber tubes in a ‘H-configuration’ allowing an easy accessible space for a 10x15x20 cm package of 1kg. The tubes are connected using laser cut multiplex wood and wings are constructed using a polystyrene-balsa sandwich structure, covered with Oracover. For a good directional stability, the wings are slightly swept-back and winglets, that also help reducing the induced drag, are added. Since the wings introduce a high moment of inertia and strong moments because of wind around the yaw-axis, the propellers are tilted 10° to improve the yaw control.
Because of the two different flight modes and the transition in between, a new control strategy is needed. This strategy contains three levels. The first level, or low level, is the angular rate control as in “Acro mode”. Because of the -90° pitch in forward flight, it becomes hard for a human pilot to control the VertiKul since a roll command results in a yawing motion and a yaw command makes the vehicle roll (in counter-intuitive direction, yaw to the left results in roll to the right!). To make the control more intuitive, a mid-level controller is designed around the angular rate controller. This controller acts as “Stabilize mode” when the VertiKul is in hover and makes an automatic transition to forward flight when a switch is turned on the transmitter. The transition to forward flight takes around 5 seconds and gradually decreases the pitch angle to build up the speed required for enough lift of the wing in forward flight. Any input from the pilot is ignored during this phase. A quaternion representation was required in order to avoid the ‘Gimbal lock’. In forward flight, the pilot inputs are only the desired altitude and heading, making it easy to fly by inexperienced pilots. Finally, the high-level controller generates a trajectory between two base stations and commands flight mode, altitude and heading to the mid-level controller.
In order to have a fully autonomous system, we also developed a docking system. The system includes an optical precision lading system, based on a PX4FLOW unit and a docking station at which a package or battery can be swapped. The VertiKul starts from one docking station with a fully charged battery and a package of 1 kg and then flies to its destination, 30km further, based on GPS. Once arrived at location, the VertiKul makes a precision landing on the docking station at that location. The battery is replaced with a full one and a new package is loaded so that the VertiKul can continue to its next destination.
The PX4FLOW camera we use for this autonomous precision landing is re-programmed in order to detect the center of the marker on the docking station and sends these coordinates to the autopilot on the VertiKul. Based on the altitude, roll and pitch angle of the VertiKul, the position of the marker is calculated and a position controller navigates the VertiKul to the landing spot. In order to be able to land at night, the marker is illuminated by leds under the surface of the translucent marker.
Check out the video here: http://youtu.be/omaxgFVDUWg
We haven’t yet been able to test the full performance of the VertiKul because of the limited test area where we can fly. During test flights we experienced a lot of influence of the wind on the big wings, making automatic landings very hard. Also the battery and package swap is not yet automated, leaving us with enough work to continue this project.
Comments
What about foldable wing ? Just an idea with performance in mind.
Thanks for all the interest and comments!
Landing accuracy with the Px4Flow is indeed within 5 cm, we wrote a paper on this topic that will be presented at 'IMAV2014" conference in Delft 4-8 August. If people are interested in the code for the automatic landing, please send me a private message. I'll make a new blog post after the conference.
Eric, 6.5kg is indeed very much for these small propellers, in hover they are pushed to the limit (18000 rpm, 250W per propeller) but the aim is to have efficient forward flight. We did test flights without payload and with a lighter battery.
Next year we will try to tackle the wind problems by decreasing the wing size and use the propellers for both forward thrust and part of the lift. A " wind protected " area is also a good solution to improve the take-off and landing
Okay, your talking about final descent in hover mode, sorry.
Well although at first I thought this is better, I'm not so sure. Because this is not a stable situation... if the wind turns the wing a bit (quadcopter yaw) it will turn in completely, since the projected surface has increased...
what I meant by "point wingtip into wind":
- we are in multicopter mode (hovering) => descending means going tail-wise with your craft
- let's say wind is coming from 90 degrees (east)
- I suggest to point "forward" of the multicopter (actually the underside of the winged aircraft) to 0 degrees (north)
Expected result: wind (coming from the side) will meet the right hand wingtip first and then pass along the wing towards the main fuselage.
Obviously, in heavy gusts of wind, this will never be perfect. But it may help to get down into the wind protected zone.
Not sure if this is understandable, maybe I should make a drawing ...
The image processing, or automated landing altogether wasn't my part, but I'll ask my collegue about it..
I didn't know about the 'red balloon finder' yet, very cool!
Thanks for the ideas about dealing with wind! However if you point the wingtip into the wind when descending you have crosswind, isn't this undesirable for landing?
Thanks Menno for the answers.
Since you mention image processing for precision landing, is the code available somewhere or do you plan to make it available? I assume you are aware of the "red balloon finder" code: http://diydrones.com/profiles/blogs/red-balloon-finder. Would love to see more people working on this.
Regarding wind: my idea would be to:
1. transition to hover in safe altitude
2. point wingtip into wind to reduce negative effects and start final descent to landing pad
3. enter a "wind protected" area around & above the landing pad for precision landing
"wind protected" area: a cylindrical / cone shaped "wall", certainly a couple of meters high & in diameter: either natural wind blockers (hedges, trees, ...) or man made wind blockers (a modification of those snow drift fences alongside roads, ...). Just an idea though.
And yes, a big downside of this concept is that it doesn't like wind. Especially for hovering and transitioning, because the wing is then perpendicular to the air stream. Indeed, maybe by aligning the wing with the airstream during landing it suffers less, this is to be tested. In forward flight wind is less of a problem.
Landing was actually a separate thesis. Unfortunately there was no time to integrate it on VertiKUL. Precision landing on the platform you see was achieved with a 450 Flamewheel and indoors. Accuracy was about 5cm, if I remember correctly, based on an imaging processing algorithm.
In hover mode it flies like AltHold, so landing was done like a normal quad (maybe switch to Stabilize just before touchdown). For forward flight a new flight mode was created. One that allows very intuitive control. You only use two sticks to control climb rate and curvature. With no input it holds altitude and heading ;-).
It accually weighs 4.8kg with 1kg payload included
Wow! Keep going guys. Make sure the next team can build on what you have created.
1. @Eric +1: what about landing accuracy?
2. How about the same in wind conditions? I presume crosswind will be the main challenge for precision landing with a winged VTOL. Maybe this can be mitigated by pointing the wingtips into wind direction?
I like your overall concept & parcel loading.