Thank you very much, Lorenzo. Yes, welcome to my presentation about extended periods of tetraflight and skysers. My name is Martin Quack. I'm in charge of flight control and skysers power. What I'm going to present today is joint work with Max and Solomon, and special announcements also go to Michael Erhard and the crew members of the

Foundation, Brave for Water. In today's talk, I will start with an introduction and give a short system overview. Then I'll go into the coordinate system that we are using, a so-called simple model. It's a set of ODEs.

Then I will talk about automated flight control for yacht traction. There I will present some data from the skysers tower, where we did a model validation using test data from our prototype.

So Skysers has been in business since 2001. We scaled up the systems from 20 square meters up to 320 square meters, going through these steps like 48 kilometers per square meter.

In 2011, research started on this power function model with a 20 kilonewton system starting with kites at 20 square meters. That's the latest data up to 40 square meters. Research on these power function models actually

included development of target point control and that has then later been ported to the Brave for Water for yacht traction. This is my last Airborne Wind Energy Conference presentation. The biggest news since the last Airborne Wind Energy

Conference is that in January 2018, Skysers started the SkyPower 100 project. The goal of that project is really to achieve an autonomous long duration twice at a larger scale. The project is supported by the Federal Ministry of Economy Affairs and Energy.

We have two power utility companies as partners and one research institute from the University of Canada. Now since I'm going to talk about the yacht traction, in the first part I should say some words about Skysers Yacht's first customer.

That is Race for Water. Race for Water is a foundation that has been founded by Swiss Marco Sibioni and the goal is to raise the awareness of plastic pollution globally and to raise this awareness day on a journey around the world using a solar vessel.

Originally it was only a purely solar powered vessel and since 2017 we added Skysers Yacht power system onto that, so now it's 100% renewable, both solar powered and wind powered. It departed 2017 from Lorient across the Atlantic,

across the city and currently as they speak they're on the way from the Philippines to London. Yes, this video should give you some impression on how the system works. We use ram air pipes to capture below the kite we have a control pod

that allows us to bring steering deflections to the system and also the control pod forces and sensors, like airspeed sensors. This is the tow point, the point where we use the traction force into the ship. Also at the tow point we actually have

line angle measurements, so we get the elevation angle after the tether. So in quick overview again, canopy control pod with steering belt, tow point and a mask that is only used for launch and landing can be folded away and this is the tow pinch

which we need to change the landing during start. So now the more theoretical part, maybe I should say this would be very quick for a 50 minute talk, but you can read it up in the 2018 Airborne Wind Energy book

and find some additional material there. So the coordinate system, what's important is it's aligned with the wind and in total we have four coordinates, three angles. Two of them are the wind-wind O angles, the wind-wind O elevation, the wind-wind O at single angle

and the third angle is actually the heading angle of the kite relative to the wind. Zero heading means heading straight into the wind. And the fourth one is the tether length. So there is an assumption on the tether, we assume that it's just a straight tether, that makes it simpler, but it allows us to

determine the tether position completely just using these two angles. So when one uses such a coordinate system, then actually the equations of motion get very simple and the main model parameters that go into that

is actually the glide ratio and the other one is the turn rate log constant which relates a steering belt deflection to change in turn rate. Now because the equations of motion are coupled and changing through the turn rate, changing the heading,

we can actually, we can reach other coordinate positions in the coordinate system. So if we speak about traction, the last equation is very simple, the cosmic light constant line length and that means we only have really one control input. There is the VA, which is the SB, the control pod,

which does not appear as a state itself, but it can compute with its kinematic relations. And also very important is the wind speed at flight level. You can either see it as a model parameter or from a control point of view, it's rather a disturbance, so then you would split it into a normal wind speed.

Now for control applications, it's very important to know which quantities can you actually measure. So the tether length, you can measure it by having a rotary folder at the towing bench and we do some bookkeeping and from that we get the tether length.

The SB, VA, we measure directed control. The wind speed, VW, and also the wind direction at flight level, it can be estimated using this SB measurement in combination with the kinematic model. The wind window angles for phi and for theta, they can be computed using these measurable

torque point angles that I mentioned before. And because the system is aligned with the wind, you also need actually the apparent wind angle at flight level to compute these. And the last one is the kite orientation. This can be estimated using a inertial measurement unit,

which is directly mounted at the control pod. So for us it's actually important to have the control pod because it forces some of the sensors. So having introduced the coordinate system and also the governing equations, we can now talk about the automatic flag for the object.

What we're using, the control scheme that we're using is target point. It has been first suggested by Doran and Leong 2013 and it has been implemented also as guidance in 2015 and I presented some adaptations to make it useful for marine vessel structures in 2017.

Just in a nutshell, the idea is if you have the kite and you have the torque point, you fly towards the torque point. At some point you will enter a trigger region which is an opportunity of the torque point and that makes the active torque point switch to the stiffness torque point. The inner loops will then take care of flying towards

this torque point, either in an 8 up fashion or an 8 down fashion. What is shown here is an 8 up fashion. And if you then look at the resulting trajectories, if you place the torque points here and here, then you will get a figure of 8.

Now, one important placement parameter for the torque points is the torque point height, or actually the wind window elevation of the torque point. And that means it's very simple. As you go lower in the wind window, that will increase the torque.

And that is shown here. If you have the torque points in red down here, then you will get a pattern like this in red and that corresponds to the red force. And that is obviously much larger than the black force which corresponds to this plug. But you cannot just place these torque points as you like.

There are some operational constraints. The first one is that you want to stick to a minimum flight altitude or you don't want the kite to hit the water surface. And that can be translated to a minimum capturing angle if you fly a constant line angle.

The other one that is very important is you don't want to rip the kite, so you want to limit the maximum tetra force. So what's important to note is that these are actually state constraints on the resulting trajectories. So you don't know a priori from where you placed the torque points

as to what the force is going to be. But you can do something like using a stepping tetra force controller. The idea is that you observe the tetra force and the tetra elevation angle and you just decrease the torque point height in small steps until you reach the desired tetra force.

Then if you're above that value it can still happen that actually through a disturbance as shown here the tetra force will suddenly go above some limit like the green line. And then the idea is to just increase the torque point height

which will decrease the force. So a controller like that will keep the force in a bandwidth between the blue and the green line. So as you see in the end you end up here again inside this bandwidth. It's a very simple approach but it's very effective.

And we handed over that controller to race the water crew in February 2018. It operated for hundreds of hours since then and it typically keeps the tetra force in a range between 14 and 20. And the main result I mean the output for the crew

is that it allows to increase the average ship speed to 6 to 8 knots using only kite propulsion compared to 4 knots when they're only using salt. So that was about the yaw fraction. In the second part I will talk about model validation

on our power control model. And again I showed three or two to show you what that looks like. So this is as we launch the system it's the same thing. We have a rare magnet control pod marked for launch and landing. But the difference is now that during the power generation phase

we are flying this figure 8 while we are wheeling out. So we're changing the line length. And as we reach at some point the maximum line length we have to switch to retraction phase where we bring the kite back in. So we actually reduce the line length. So in the model that means actually

everything else still stays the same but you get here the different equation which is you just have the wing speed as the controller. So if we talk about model validation the question is can the model describe the system's behavior accurately?

And maybe also can we make predictions into the future using that model? So if we want to do such a model validation there's two possible approaches. So you typically use for both approaches you typically start with some measure of data from the tri-test. Let's call this measurement data y and you have also some recorded

control inputs u. Now the open loop point of view or the open loop approaches that you just take this data and you pick an initial point x0 and then you integrate the equations of motion into the future and you need to use some control inputs

you just interpolate control inputs from the control history during the experiment and that will be the open loop point. Now the other way you can go is you don't use these recorded control inputs but you actually regenerate them in the simulation by just

re-running the control loops using exactly the same settings as you used during the flight test. And there is another thing that is the wing speed at flight altitude that is always a bit disturbing because we cannot measure it. However, we can measure the air speed and we can then

estimate. To show you in this time frame that the power cycle is going to be which I show in the next slide we have in red the wind speed at ground. It varies between 5.5 and 10 meters per second. And then we have in black the estimated wind speed at flight altitude

which is definitely larger in average than ground. But don't be irritated the smoothness of this black line that is rather than a result of the estimation algorithm. And probably not as smooth as the black line. And then in blue is the wind speed

value that we used in the closed loop validation. So if you look at the results these are the total states. And first for the open loop simulation one thing that we noticed, well the line length matches very nicely because it's very simple. We just integrated the control. But then if you look at the elevation angle

we actually have a deviation. So, but for the other states it looks good. But it breaks away. But as we do a closed loop simulation we don't get any deviation in the elevation angle. And also the other

values they at least stay in bounded. The line length is not as nice for this because we also do have some disturbances in the flight test while not doing the simulation. So you get the right picture in average using the closed loop simulation.

And that can be seen better if you look at flight trajectories. So the flight trajectories actually match quite nicely if you plot them correctly. So in summary for the first part automatic flight controls for marine traction are hundreds of light levels that have been collected

in a powerful automated frost wind kite flight. We increased the average recipe to six way announcement on the kite. And I think a very important point is just really be handed to the crew who knows no people, no entities from the skies at what time they are using the system. And that means that we need to reach a very high

technology readiness level. The simple model from the model validation in the second part as we get the average picture correctly, I think it's fair to say that it describes the system behavior adequately. And also I think that predictions seem possible using that model.

But they are definitely subject to uncertainties in wind at the flight altitude. So I think that opens the opportunity to MPC or the use of non-linear model predictive control and maybe makes a nice link to the very first presentation.

And I think that these approaches they actually provide the pathway for the automation codes that we set inside the SkyPower model project. With this I would like to close the presentation showing the acknowledgement slide and there is one more thing. We are actually hiring. So visit us at the control booth and

check our web page. We are currently also looking at control engineers. So I hope there are some control engineers in the control session. So with this I would like to close the session.