Technology and Calculations

How to design a flying Exoskeleton

Over the years, we have assembled a lot of information and calculations for the design of the flying Exoskeleton. Have a look at Human Birdflight for a historical introduction and a glimpse at the scientific literature. On our journey we faced many questions like: Does it even fly? How large are the wings? How heavy is it? Can a human control this thing? Is the available material strong enough?

This page shall dive deep into the topic and shed some light on the hard facts and calculations. It will not capture the topic fully, though, as there is much too much to put it here. Nevertheless, it shall provide a forward base for further developments in the future. Some of the most critical and fundamental calculations of our basic design study shall be presented here.

This article requires some prior knowledge and is geared towards engineers.

For literature citations see human birdflight feasibility

Preliminary Thoughs

The basic idea, as described in the patent, is to let people fly like a bird through robotic wings attached to a backpack. The wings are controlled with the arms/body via an exoskeleton and return the aeroelastic forces back to the pilot via the exoskeleton. Thus, the pilot can feel the air under the wing and "learn to fly" in the best possible sense. No direct mechanical connection between the control input (e.g. the arms) and the output at the aerodynamic surfaces (e.g. the wings) should exist. They are coupled via the flight control computer/the teleoperation system. This way, the amount of feedback force on the arms and legs/ankles can be adjusted or even turned to zero. The pilot can decide how much feedback is desirable for a good experience.

Further necessary parts of the system are, a power source to drive the robotic actuators and other subsystems, tail robot for aerodynamic control, and safety/emergency systems.

The system will thus consist of:

• Pilot
• Exoskeleton
• Backpack housing support subsystems
• Wing robots attached to the backpack
• Tail wing robot

To generate lift for flight, two options exist analogous to birds. Both options correlate with distinct flight modes:

1. Gliding and flapping flight. Steady forward movement generates dynamic lift at the wings, which requires a minimum true airspeed. Thrust can be generated by flapping the wings.
2. Fluttering flight. Unsteady dynamic lift generated by wing flapping. Potentially using unsteady aerodynamic effects. Does not require motion of the whole system relative to the air and is thus usable for vertical take-off and landing (VTOL) and hovering.

In the final installment, the system is supposed to be VTOL-capable (just like birds). So it will need to be able to lift its own weight with the thrust generated by fluttering. As this mode drains the most power and puts the highest strains on the mechanical structure, it can be considered a system driver of the final system. Both modes will be investigated in the following.

Design process summary

The process for the preliminary design study is as follows. First, the system's basic dimensions necessary for gliding are determined. Gliding is the first basic flight mode used in steady flight and landing. The process has these steps:

1. Have a rough idea of the system. (see e.g. here)
2. Estimate the weight of the components/structure plus the pilot.
3. Estimate the wing geometry and maximum lift coefficient.
4. Estimate necessary wing area with weight, $c_d$, and a desired stall speed.

With the basic questions answered, further flight modes can be investigated. For propulsive flapping flight, the process is:

1. Decide for a maximum allowable flapping frequency.
2. Decide for a maximum flapping amplitude.
3. Calculate thrust, aerodynamic efficiency, optimum wing torsion using the research of Dr. Send. For hints on understanding the paper see below and aniprop.de. For an excel implementation of the paper see berblinger2.de and "Physik der schwingenden Platte" at aniprop.de. Dr. Send's body of work can be found here

Further questions can then be answered:

• What moments and forces are necessary for fluttering?
• Are there materials and acutators capable of handling these?
• What is the aerodynamic efficiency (gliding angle) of the system?
• How to acheive aerodynamic control and stability?
• What are the power requirements of flappling and fluttering flight? What systems can provide this?
• How long can the system stay in the air? What is the maximum range? How high can it fly?
• Is the above made assumption of maximum weight reasonable?
• And are humans fast enough to control the system, or is kybernetic augmentation necessary?

Answering these questions can get arbitrarily complex, considering all the different load cases, missions profiles, degrees of freedom, and geometric designs the system might attain. For a preliminary study of feasibility, we had to somewhat limit ourselves.

Does it fly? - Dimensioning the Wings

Bird wings use dynamic lift of an airfoil to enable flight. Lilienthal figured this out and built some of the first commercially available airplanes of the world. The lift is calculated with:

with the lift force $L$, the lift coefficient $C_L$, the air density $\rho$, the true airspeed $u$ and the wing area $S$.

As the sharp engineer can see from these equations: even a rock flys if you stick wings to it. The question is just: "how well?" The better the system aerodynamics, the bigger the wings, the higher the fluid density, and the faster you go, the more lift is generated.

How much lift? - Gliding

How much lift is needed depends on the weight of the whole system including pilot. As the details of the system are not known yet, we have to estimate. The pilot is assumed to weigh $80\,\mathrm{kg}$. For the system we firstly assumed a maximum of $60\,\mathrm{kg}$, as this is roughly what a well-trained person can shoulder. So a total minimum lift of $L = G = 140\,\mathrm{kg} \cdot 9.81\,\frac{\mathrm{m}}{\mathrm{s^{2}}} = 1374\,\mathrm{N}$ is requied for flight. Of course a more light-weight system would be highly desirable. But making it lighter requires more engineering. This is just a first shot and wieght reduction will happen over time. And the idea is anyways to start somewhere and then iterate the design. After finding geometry, forces and some exemplary parts, we did a rough check on our initial assumption of $60\,\mathrm{kg}$ for the whole system and it seemed reasonable.

$C_L$ and $S$

Now that we know the ballpark of the required lift, over which path can we generate it? As the system is unknown its aerodynamics expressed in $C_L$, are also unknown. An estimate for the profile lift coefficient $c_L$ not higher than 1.5. Birds can acheive $c_L$ of up to 4 by using unsteady aerodynamic effects. And the ground effect will substantially boost lift near the ground (duh..). But for this first draft, we'll stick to conservative estimates from airplanes and hang-gliders. The profile lift coefficient is not yet a usable number due to the finite length of the wing and a consequential wingtip vortex. To estimate $C_L$, one has to first define some wing dimensions.

Let's assume a trapezoid wing with a span $b$ of $12\,\mathrm{m}$, cord length $r$ of $1.5\,\mathrm{m}$ and $\delta$ reduction to 0.3. The resulting wing area is gained by

With the rough geometry defined, one can estimate $C_L$ for the finite wingspan. Wing stretch is gained by

$C_L$ is then estimated with:

Stall speed $u_{min}$

With all necessary values estimated, the stall speed can be calculated from the above lift equation to be $u_{min} = 10.9\,\frac{\mathrm{m}}{\mathrm{s}} = 39.2\,\frac{\mathrm{km}}{\mathrm{h}}$. This might seem a lot to run with $60\,\mathrm{kg}$ on the back. But keep in mind that the ground effect will cut this down to size. The values for $S$ and $C_L$ correspond to those of hang gliders, which are quite capable of landing and starting with a similar weight (much heavier pilot, though). Also, just like birds, the system is supposed to be able to flutter. That's a different flight mode though, and will be investigated below.

Curve radius

For the concept of a stadium sport, the minimum turning radius is important, as the space is limited. This directly follows from u_min, when a maximum g-force is defined via the centripetal force. We'll assume a conservative estimate of a 3g-turn. The radius then follows as:

With common soccer fields having a size of $100\,\mathrm{m}$ times $80\,\mathrm{m}$, this seams feasable with a large margin. N.B.: we are aware that this says nothing about corresponding aerodynamics. With a wingspan of $12\,\mathrm{m}$, a turn of $4\,\mathrm{m}$ seems ambitious, with parts of the inside wings not seeing reversed flow...

Wing root moment - gliding

Let's have a quick look onto what kind of moment we would have to generate at the wing root, in order to hold the wing in position. We'll assume an elliptical lift distribution. Birds actually seem to use more of a bell-shaped lift distribution with small negative lift on the wing tip due to twist. But the elliptical distribution yields a more conservative estimate. Assuming a double trapezoid wing, the calculationand a double trapezoid wing, the calculation of the lever of $2.4\,\mathrm{m}$ (see details in the calculation package here).

With half of the weight per wing, one arrives at a root moment of $M_0 = 1648\,\mathrm{N\,m}$ for gliding. This fairly large moment will define the main actuator of the wing actuator system. All other actuators can likely be significantly smaller. Indeed, when looking at the reference of birds, the breast is huge in comparison to all other muscles.

Actuators to produce the root moment

We investigated three different types of possible actors. The first are piezoelectric polymer fibers, which are an interesting concept and have been used in some exoskeleton systems. For the main wing actuators, the force density is too low, though. They might be an interesting option for a soft exoskeleton or other low-force applications.

The second option are small hydraulic cylinders. With a cantilever mechanism and $150\,\mathrm{bar}$ pressure, a cylinder of $2.6\,\mathrm{cm}$ diameter is required. This works in principle with existing technology. The drawback is the large oil volume necessary for the whole system. This alone would add about $10\,\mathrm{kg}$ of weight for the main actuators. The actual cylinder and other systems are not the main weight constituent here, as there are versions made of carbon fibre available. Due to the weight, this option seems undesirable.

The third option is electromotors and a transmission system. Motors of a desirable size, weight and power are commercially available (e.g. HACKER Motor Q100-6M kv168 with $9.3\,\mathrm{kW}$ peak power). They require a good amount of transmission in order to not get too big and heavy. We assumed a transmission ratio of about $100{:}1$. This transmission can be realized using a harmonic drive (or a cycloidal drive) or a multi-stage belt-and-pulley system. The former is relativeley heavy, so a belt system is probably desirable.

The pictures below are also available in larger format here

Structure

Most of the structure will be made of glass and carbon fibre composites for their lightness and strength. Parts requiring isotropic or very hard materials will be made from metals (Titanium, Aluminium, hardend Steel). The wings are supposed to be partitioned in three, just like bird wings. The parts are named accordingly "humerus" for the "upper arm bone", "radius" or "ulna" for the "lower arm bone", and "digitus" for the "hand" structure. In birds, the hand is usually reduced to:

• a thumb remnant, often controlling the thumb feather as a kind of nose-flap,
• a prominent "index finger" controlling the hand feathers. In birds, the relatively short humerus protrudes from the shoulder in a backwards sweep. At the elbow, the leading edge angles in the upstream direction. Lastly, at the wrist joint, the short digitus bone and hand feathers sweep backwards again. Most of the outer wing is actually just feathers without an internal bone structure. Probably, because this allows for the nice aerodynamics of the hand wing in flapping and fluttering flight. But also because it keeps the inertia of the wing small during flapping/fluttering and is thus energy efficient. The length ratios of humerus, radius, and hand wing vary between bird species. Usually, humerus and radius make up about 40 percent of the wing length while the hand wing makes up the rest (see data on birds here). Thus, the bone-less hand section encompasses a sizable portion of the total wingspan.

A general remark of Berg and Rayner on the wing design: "The best predictor for the moment of inertia is the product of wing mass and the square of wing length: $I=0.118(m_w l_w^2)^{1.04}\ (r^2 =0.997)$ (Fig. 3B; Table 3). This suggests that the mass distribution of the wings of each bird species reflects a similar compromise between reduction in the cost of beating the wings and maintaining a reasonable stiffness and strength. ($m_w$: wing mass, $l_w$: wing length)" Source: Berg, Rayner - 'The moment of inertia of bird wings and the inertial power requirement for flapping flight'; The Journal of Experimental Biology 198, 1655–1664 (1995)

There are a large number of possible loading conditions resulting from the many degrees of freedom, flight modes, and mission profiles. With our small team of engineers, we had to limit ourself to a rudimentary feasibility analysis. Thus, we only estimated, if the largest required forces can be handled by a CFRP beam. That turned out fine. But for some reason, the docuemnt with that calculation got lost (sorry!). And I'm too busy now to do it again (not sorry). I only could find some data we used for the calculation. So if you're up for it, help yourself:

• Fibre-volumefraction: 50% (Vacuum infusion)
• Epoxyde resin density: $1.15\,\frac{\mathrm{g}}{\mathrm{cm^{3}}}$
• Carbon fibre density: $1.77\,\frac{\mathrm{g}}{\mathrm{cm^{3}}}$
• Tensile strength and modulus laminate: $820\,\mathrm{MPa}$, $60\,\mathrm{GPa}$
• Compression strength and modulus laminate: $635\,\mathrm{MPa}$, $54\,\mathrm{GPa}$ The data for laminate strength is from a fibre/fabric manufacturer's data sheet and is likely very conservative. For reference: the tensile strength of the fibres is stated to be $3750\,\mathrm{MPa}$.

Thrust - powered flight

In difference to airplanes, birds generate thrust by flapping and morphing their wings. The aerodynamics of this process are necessarily unsteady. It begs the questions: how is the movement pattern supposed to look like? How fast has the wing to flap and with which amplitude?

To understand the basic ways birds move it is useful to study the literature and slow motion videos. In the mode of steady forward flight, birds flap their stretched-out wings roughly $60^{\circ}$ up and down at max. For long distance flights, this amplitude is lower. Presumably to save energy. During the downstroke, the wing is twisted forward such that the wing hand (pinion feathers) is turned in the direction of the downward movement. This turns the lift vector forward, producing thrust. During the upstroke, the twist is in the opposite direction, albeit less pronounced. If moved quick enough, the outer wing part actually has inflow coming from slightly above, which results in thrust and downforce on the outer wing part. Birds also move their wings forward and backwards (w.r.t. the direction of flow) during the stroke. This is not strictly necessary to generate thrust, though, and will be ignored for now.

Flapping Frequency - Empirical Research

How fast should the wings beat? Can a human facilitate that?

To gain an initial estimate of frequency $f$, one can look at scaling laws from the literature on birds. Berg and Rayner stated in 1995 two laws:

• $f \sim b^{-2/3}$
• $f = 1.807\ l_w^{-1.055}$

With the wingspan $b$ and the wing length $l_w$. The exponent of the latter equation actually had a 95% fiduciary interval from $-1.29$ to $-0.86$. Assuming a wingspan of $12\,\mathrm{m}$, the frequencies become $0.19\,\mathrm{Hz}$ for the first equation and $0.2729\,\mathrm{Hz}$ ($0.179\,\mathrm{Hz}$–$0.387\,\mathrm{Hz}$) for the second.

Another scaling law is by Greenwalt from 1962: $f \sim l_w^{-1.15}$. This yields a frequency of $0.127\,\mathrm{Hz}$.

These frequencies are quite attainable by humans flapping their arms.

Theory of Thrust Generation

The correct relative amplitudes and phase angles of the flapping movement can be derived using potential theory. This has been worked out since 1935 when Theodorsen wrote his paper on flutter. In 1992 Dr. Wolfgang Send picked up this thread. His work describes a 2-dimensional profile undergoing plunging and rotating motion with potential theory. He calls this the bending/torsional drive. Roughly speaking, bending means here the flapping of the wing around its root, while torsion means the changes of the angle of incidence of the wing(part). The theory was successfully applied in the Festo Smart Bird.

The upshot of the theory is that the wing tip twist has a phase lag of approximately $90^{\circ}$ with respect to the flapping. So when the wing starts to flap downwards, the chord is horizontal. Half of the downstroke the wing tip points downwards. And at the end of the downstroke, the chord is horizontal again.

As the flow at the wing changes with flapping frequency and amplitude. In order to achieve thrust, the correct movement pattern has to be executed to get the correct incident flow.

Movement Parameters and Dimensionless Numbers

Let's assume first, we would first choose a flapping frequency $f = 1.2\,\mathrm{Hz}$ (because that is about the maximum what we could do with our arms). Then a parameter called the reduced frequency $\omega^*$ can be calculated as follows:

where $c = S/b$ is the average chord length, and $u$ the free stream velocity. This number relates the downward motion to the forward motion. If this motion is supposed to be aerodynamically efficient and produce thrust, the flapping speed and wing twist have to fit together. This is captured by the approximated relation of the figure of merit $\eta$:

with ratio of amplitudes $\lambda$. This ratio connects the amplitudes of flapping and twisting. It is defined as:

where $h_0$ is the flapping amplitude, $\alpha_0$ is the twisting amplitude

If one assumes somewhat efficient thrust production and thus chooses some value for $\eta$ (e.g. 0.75), the ratio of amplitudes can be calculated ($\lambda = 5.56$). And from the ratio of amplitudes, the twisting amplitude $\alpha_0$ can be calculated, given a flapping amplitude $h_0$.

$\alpha_0 \cdot c/2$ is the arc length, of the wing twist at a radius of c/2. $h_0$ also can be considered an arc length, and thus can be calculated from the deflection angle ($\pi/6 \mathrel{\hat=} 30^{\circ}$) times the radius of the representative wing profile (2/3 wing length). The resulting angle of the wing twist is even higher than the flapping angle, which seems counter-intuitive. But keep in mind that for aerodynamic velocities, not the angle is important but the distance traversed.

Caveats

As the theory is derived in 2D, transferring the concept to a 3 dimensional wing has pitfalls. Due to kninematics, every radius yields a different $h_0$. For aerodynamically optimal thrust generation, the resulting $\alpha_0$ would need to be gradual along the wing. In our technical system, the wing twist is not gradual, but only realized by rotating the outer wing part. Thus, a representative radius is chosen, for which the angle $\alpha_0$ is calculated. Dr. Send also analyses in his 1992 paper a 3D version of his theory and remarks that looking at the flow section by section is an acceptable approach.

Further on, the flapping amplitude can be chosen freely (to a certain extent). In Dr. Send's model, the approximation of small angles (e.g. $\tan(\alpha) \propto \alpha$) has been used in a few places. Large angles can and will therefore lead to inaccuracies.

Also, choosing an aerodynamic efficiency as input does not mean, that this efficiency will actually be reached. This will have to be tested. In his work, Dr. Send described also that the best efficiency actually was reached with a slightly larger phase lag of $95^{\circ}$–$100^{\circ}$.

How much Thrust can be produced?

Not much could be found on this, but in one place Dr. Send gives the formula:

with the average (Thrust-)Force $\bar{F}$, the drag $F_D$, the phase lag of twisting to flapping $\kappa$, and the maximum angle of the inflow $\alpha_{H_0} = \frac{h_0\cdot 2\pi\cdot f}{u}$.

Dr. Send writes: "The coupled flapping and rotational motion of a wing is capable of overcoming the steady-state drag generated by the wing when the angle of the instantaneous inflow due to the flapping motion is greater than the angle due to the rotational motion. The mechanism is particularly effective when the flapping motion leads the rotational motion by $90^{\circ}$."

It makes sense that something like the "incidence angle" between the flapping wing and the instantaneous inflow dictates thrust. But it is also insteresting, that the wing twist $\alpha_0$ impacts the thrust quadratically. Presumably, because the more the wing is twisted, the stronger the lift vector is rotated into the flight direction.

Fluttering flight

This flight mode is somewhat different to normal flapping. The flapping angle becomes quite large and forward motion of the system is small. Most of the inflow on the wing is produced by the active flapping.

In this mode, effects of unsteady aerodynamics are very pronounced. The periodic and strong change of angle of attack produces vortex interactions behind the wing that change the flow at the wing. Bumblebees and Hummingbirds use this effect very strongly, basically pushing themselves off their shed vortices. In fluttering, stall is delayed, as disturbances in the flow don't necessarily have enough time to traverse the airfoil fast enough to separate the flow. This can yield very high lift coefficients of up to 4. If the tips of both wings get very close when the down-/upstroke start, a kind of sucking effect can be induced between the wings. This yields extra thrust and reduces the starting vortex shed.

Larger birds usually lift off in a rather shallow angle. Probably due to power limitations of the flight muscles. When medium sized birds lift off (VTOL), they use a different flight pattern. The wrist and wing folding becomes important. On the backstroke they fold the inner wing (shoulder and elbow) and tilt their hand backwards up to $90^{\circ}$ to the wing surface. This allows them to push air downwards during the backstroke, increasing the lift produced. See e.g. this The small feathers on the back of the wing also help prevent stall in this extreme flight mode, as they prevent flow inversion by lifting a little off the surface. This can be very well seen in this slow motion footage. Here is some slow motion footage by the BBC and the Slow Mo Guys:

Due to all of these strange effects, thrust generation and power requirements of fluttering is hard to calculate. We made a (probably wrong) attempt that didn't consider any of the complex aerodynamics but just the rotation of the wing. This only delivered a possibly useful value for the working point of the lift force (3/4 of the wing length). You can find the calculation here.

Power requirements & power systems

To supply enough power to the system, several options exist. For powered flapping flight, a larger LiPo Battery might suffice. In this flight mode, mainly potential energy gain and drag have to be compensated. The more energy hungry fluttering flight can be supplied with small shaft power engines ("Turboprops") with about $9\,\mathrm{kW}$, weighing $2.9\,\mathrm{kg}$ per engine (e.g. Jetcat SPT10). They are similar to what Gravity uses to propel its jet suit. The lift generation of the flying exoskeleton is a lot more efficient, though, and can provide much longer flying times per litre fuel.

The ballpark of power requirements can also be estimated from the examples of flapping flyers. There exists a large body of research on power requirements of birds, bats, moths and even pre-historic flyers. One example from the last category is Quetzalcoatlus Northropi, the largest known Pterodactyl dinosaur. Data for those beasts from the lost world are obviously questionable. But it is estimated to have lifted $250\,\mathrm{kg}$ into the air, having a wingspan of $10$ to $11\,\mathrm{m}$. From the standpoint of our own analysis, the resulting wing loading seems a bit much, leading to very high stall speeds and thus, landings. Maybe these giants could only land by diving into bodies of water and lift off from cliffs. In any case, the power requirement for lift-off was estimated by Marden in 1994 to be about $13\,\mathrm{kW}$. Birds nowadays seem to require about $20$ to $50\,\frac{\mathrm{W}}{\mathrm{kg}}$ bodyweight (Tobalske 2003). This would mean a power requirement of $7\,\mathrm{kW}$ for the Flying Exoskeleton, assuming the upper bound. Another estimate for the minimum required power, based on the mass of the flyer, arrives at $3.5\,\mathrm{kW}$ for sustained flight (scaling law by J.M.V. Rayner 1979).

Using aerodynamic reference values from hang gliders, one can also estimate the power required for sustained forward flight. The glide ratio $\epsilon = C_a/C_w$ of those flyers is about 15. One can assume a $C_a$ of 0.75 for distance flight (s. e.g. Berblinger2 paper by Dr. Send). The power reqired then follows from the drag force:

This further assumes a velocity $u$ of $20\,\frac{\mathrm{m}}{\mathrm{s}}$, the wing area $S$ from above and an air density $\rho$ of $1.22\,\frac{\mathrm{kg}}{\mathrm{m^{3}}}$.

Machines can have a much higher force- and powerdensity then biological systems. Applying even $20\,\mathrm{kW}$ of power is not a real issue when using tubo engines. On the other hand, birds and other small flyers can profit more from unsteady aerodynamics, as they move in a different Reynolds-number regime. Also, bird feathers are still hard to reproduce technically (aerodynamics + stiffness). So the Flying Exoskeleton does not necessarily outperform birds off the bat. In any case, these estimated power requirements seem attainable.

Control

Control can be an issue with a bird-like tail, as it was during the development of the SmartBird. As the arms are busy and for the authentic feel, this tail should be controlled with the feet. The exoskeleton at the ankles can detect the angle and transform this into control output to the tail. Bird tails do not split elevator and rudder, but have one aerodynamic surface for both purposes. The tail can rotate along its axis and bob. As a result, the tail can produce moments in any direction around the body center. Birds actually often don't use the tail like a typical elevator, but regulate tilting moment via wing twist. The tail is sometimes even used to create extra lift.

Controlling roll via the arms, pitch and yaw with the feet has been tested by Fluglich using the Exoskid and the Exoskid Trainer. This distribution was indeed found to be intuitive, even though tail control via the feet requires some training. In our setup, the rudder was controlled by the angle-difference of the ankle joints, while the elevator was controlled by the average angle of the two joints.

Safety Systems

Despite what every second German out there seems to believe, we have also spent some thoughts on safety. From pilot error or system failure stems arisk of crashing.

When far enough above ground or in stable flight

The wings are supposed to be designed in a way, that they return to a neurally gliding configuration upon power loss. This can be facilitated by wires along the core of the wing skeleton. When power is lost, these wires are tensioned by springs. The wing joints can be designed in a way that they return and be held in the desired position this way. Alternatively there could be inflatable bags inside the joints that get filled by pressurized air cartridges on power loss, acheiving the same effect.

Even with one wing broken or ripped off, the system would still rotate down to the ground like a maple seed. A possible option would be to separate backpack and exoskeleton, ejecting the pilot with a parachute. This can work even relatively close to the ground, if an explosively inflated chute is used (has been only used in drones up to now).

Close to the ground

This part is more tricky. If the system is still reactive, the wings could be used to cushion the crash, destroying the wings, but saving the pilot. Other than that, the exoskeleton could contain inflatable "airbags" around the body, forming a full body airbag.

General safety systems

The flight computers and sensors are supposed to be redundant as in commercial airplanes. But many of the critical componens, like the power electronics, can not be made redundant due to weight constraints.

Reaching above the helmet from the backpack there is supposed to be a crash bar protecting the neck in case of a belly landing. The backpack will probably need to contain a fire control system.

Designing the Exoskeleton

The exoskeleton is the human-machine-interface of the system. It has several tasks:

• measuring and delivering position and forces at the limbs
• supporting the legs and the arms against gravity's pull
• protecting the pilot from cold and other physical harms
• providing a flight information system
• implementing matters of design

In its most sophisticated form, the exoskeleton will consist of 7 degrees of freedom (DOF) for each arm, and 3 DOF for each leg. The arms are directly mapped to the DOF of the wings, allowing to control the following angles:

• dihedral angle (flapping, shoulder 1),
• angle of attack (shoulder 2),
• wing sweep angle (shoulder 3),
• folding the wing elbow (elbow),
• folding the wing hand (wrist 1),
• dihedral angle of the hand (wrist 2),
• angle of attack of the hand (aileron equivalent, lower arm).

The DOF of the ankle controls the tail, as described above. The two additional DOF are used for weight offloading through a lower-limb exoskeleton (knee and hip).

Not all of these angles are necessary for the first prototype. The minimum setup controlling horizontal flapping flight only requires aileron, tail and flapping. This has been realized with the Exoskid.

Soft or Hard?

If the exoskeleton should be hard or soft, is a matter of debate. The soft exoskeleton uses the bones of the body as structure and only applies forces via pulleys. This has the advantage of being lightweight. However it has the disadvantage of being imprecise in measuring and applying force and position. This can make it hard to apply proper control theory. The hard exoskeleton has a structure of its own. It is easier to approach from a control theory side and can be implemented with common engineering tools and materials. Its main disadvantage is biomechanics. As the structure is external to the body, it has to follow the movement of the limbs relatively faithfully in order to not hurt the operator. This can be a significant challenge, especially with the shoulder joint. Nevertheless, due to the ease of manufacture and instrumentation, Fluglicht has decided to build a hard exoskeleton.

Helmet

For navigation and system control, the pilot will wear a helmet with built-in Head-Up-Display in the visor. This will provide typical altitude, attitude, orientation, speed, turning rate, etc. data as known from airplane instrumentation. Perception of movement could also be made available in an intuitive way by virtually projecting an array of dots into the surrounding volume via the visor. This way, the pilot can comprehend movement relative to this dot grid.

Technological Next Steps

• Create a CAD model for the full-scale tethered wings prototype and build it. This doesn't need to have an exact aerodynamic and structural design.
• Create realistic/producible design for the exoskeleton and build it.
• Research question: how does a teleoperation system behave, where there is a mechanical connection between master and slave? This feedback route is out of scope for typical teleoperation.
• Exoskid: giving the person a feeling for where "zero deflection" is, is important as long as there is no force feedback from the aerodynamic forces.

How to cite this article

P. Jüstel - "How to design a flying Exoskeleton", Febuary 2026, https://fluglicht.com/blog/technology_details