You are on page 1of 10

Aerospace Science and Technology 8 (2004) 479488

www.elsevier.com/locate/aescte

Study on the relative motion of parafoil-load-systems


Untersuchung der Relativbewegung von Gleitfallschirm-Last-Systemen
Gordon Strickert 1
DLR, German Aerospace Center, Institute of Flight Systems, Lilienthalplatz 7, D-38018 Braunschweig, Germany
Received 18 August 2003; received in revised form 19 April 2004; accepted 19 April 2004
Available online 12 May 2004

Abstract
Based on the experimental vehicle ALEX the relative motion between parafoil canopy and payload is surveyed. This motion results from
the flexible textile linkage of parafoil and load. Specific movement patterns are excited by maneuvers or gusts. The study characterizes the
different types of relative motions and their impact on the flight qualities of the vehicle. For modeling purposes, these additional degrees
of freedom are taken into consideration. A sensor system is designed and implemented for recording the relative motion during flight tests.
Flight data is acquired with a customized video-measurement system combined with image processing algorithms (tracking). In a further
step, the parafoil-load-system is modeled with respect to the multi-body-problem and the aerodynamic effects in a computer simulation
environment. By means of system identification techniques the parameters of the models are estimated and then validated using additional
flight test data. These computer models are used for simulation studies to analyze the effects and impacts of relative motion. Also, with
this extra knowledge both measurement improvements and deeper insights to the nature of flight mechanics are obtained. This leads to the
concluding discussion of the applicability of the proposed methods for future projects.
2004 Elsevier SAS. All rights reserved.
Zusammenfassung
Am Beispiel des Experimentalsystems ALEX wird die Relativbewegung zwischen Gleitfallschirmkappe und Nutzlastkapsel untersucht.
Diese Bewegung wird dadurch ermglicht, dass die beteiligten Krper nicht starr, sondern flexibel mittels textiler Strukturen verbunden
sind. Dadurch werden zustzliche Bewegungsfreiheitsgrade in das System eingebracht, die entsprechenden Bewegungsformen werden durch
Manver oder Ben angeregt. Die unterschiedlichen Relativbewegungen werden zunchst definiert und auf deren Implikationen eingegangen.
Es schliet sich die Auslegung und Implementierung eines Sensorsystems an, welches in der Lage ist, die Relativbewegung im Fluge
aufzuzeichnen und zu vermessen. Dafr kommen Methoden der Videodatenanalyse (Tracking) sowie der Signalverarbeitung zum Einsatz.
Auerdem werden Simulationsmodelle der Relativbewegung aufgebaut, welche neben Mehrkrperaspekten auch aerodynamische bzw.
flugmechanische Effekte bercksichtigen. Durch Verfahren der Systemidentifizierung werden das Modell und dessen Parameter angepasst.
Mittels weiterer Flugversuchsdaten erfolgt die Validierung der Modelle, so dass diese fr Simulationsstudien genutzt werden knnen.
Ergebnis dieser Studien sind Aufschlsse ber die Einflussgren und Auswirkungen der Relativbewegung. Die Erkenntnisse flieen ein in
Vorschlge zur Bercksichtigung der Relativbewegung bei flugmechanischen bzw. messtechnischen Fragestellungen. Sowohl die Ergebnisse
als auch die entwickelten Verfahren werden hinsichtlich ihrer bertragbarkeit auf andere Probleme diskutiert.
2004 Elsevier SAS. All rights reserved.
Keywords: Motion analysis; Video tracking; System identification; Aerodynamic and multi-body co-simulation
Schlsselwrter: Videodatenbasierte Bewegungsanalyse; Systemidentifizierung; Multidisziplinre Simulation; Mehrkrpersysteme

1. Introduction

E-mail address: gordon.strickert@dlr.de (G. Strickert).


1 Phone: 49 531 0531 295 3266, Fax: 49 531 0531 295 2647.

1270-9638/$ see front matter 2004 Elsevier SAS. All rights reserved.
doi:10.1016/j.ast.2004.04.003

Parafoil-load-systems are unique devices for recovery,


rescue and delivery of air cargo. Known since the late 60s

480

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

they obtain importance for todays applications because of


several reasons:
They are cheap, or better say low-cost, compared to
fixed wings.
They are lightweight and small (when packed), crucial
for aerial delivery and space recovery.
Its possible to face wind-offsets because of their ability
to glide and steer. This is a major advantage against
conventional (round) parachutes.
Precise and very soft landings are possible, which
makes parafoil-load-systems applicable for sensitive
instruments or injured humans.
In science and industry there are two major efforts for the
enhancement of these systems made: to scale the parafoils
up for higher payloads, and to add autonomy to the systems, enabling automatic guidance, navigation and control
(GNC) tasks. The Institute of Flight Systems of the DLR
Braunschweig focussed on the latter task and developed an
experimental vehicle called ALEX for the demonstration of
an autonomous landing of a parafoil-load-system [4,5]. The
design of a GNC-concept presupposes fundamental knowledge of the systems qualities and performance. A flight mechanical model is required which predicts the reactions to
control inputs and disturbances. Primary for the development of this model are flight tests in which excitations and
significant states of the system are recorded, providing data
for a supplementary analysis.
In this procedure the modeling of the relative motion, i.e.
the movement between canopy and load, is a task of special
interest. On one hand quantitative information of that motion
is needed to convert data coming from sensors mounted to
the load to corresponding values for the canopy. On the
other hand relative motion is an integral part of the flight
mechanical model itself.
Usually models for parafoil-load-systems neglect relative
motion. It is supposed that canopy and load are linked stiffly
and that there is no interference between relative motion and
flight qualities. Since the connection between canopy and
load consists of flexible harness and lines and the canopy
is not rigid, by this simplification a number of degrees of
freedom are ignored.
These degrees of freedom are to be considered by an
extended model of dynamics. Due to its complexity the
formulation of that model does not result from analytical
methods but on using a multi-body simulation tool. It is
meant to map geometry and degrees of freedom of the real
ALEX-system as best as possible to allow a comparison
between measured and simulated data. Since the excitation
of the relative motion happens by aerodynamic forces,
the model of dynamics has to be integrated into a flight
mechanical simulation environment. This over-all model
is embedded into another software-tool to perform the
parameter identification. Thus an automatic adjustment of

model parameter is supported, aiming at a best possible


match between modeled and measured behavior.

2. The ALEX-system
ALEX is a small Autonomous Landing Experiment.
Its parafoil is an of-the-shelf device like it is used for
sport purposes. Instead of the jumper a capsule of app.
110 kg weight is fixed beneath the canopy with a special
harness. This capsule carries all the sensors necessary for
the recording of flight test data. Supplementary actuators are
mounted which operate the control lines for the breaks of the
parafoil. Commands can be given by radio-control or by the
on-board-computer [5].
In a flight test the capsule is dropped by a helicopter from
a height of 6002000 m. After a short free fall phase, the
parafoil opens automatically; maneuvers or other tests can
be conducted.
Further details are listed in Table 1 and Fig. 1.
Table 1
ALEX specifications
Canopy

Capsule

Performance

Type

Parafoil 252-7

area
span
length
width
mass
lift to drag
flight velocity
sink rate

20 m2
5.7 m
1250 mm
500 mm
110 kg
<2
10 m/s
5 m/s

Fig. 1. ALEX-system.

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

481

Fig. 2. Relative yawing.

Fig. 4. Lateral displacement.

Fig. 3. Relative rolling.

3. Relative motion
The parafoil canopy, the harness, and the suspension
lines all consist of textile, flexible materials. This flexibility
introduces additional degrees of freedom for the possible
motions of the parafoil-load-system [9]. These motions are
not only theoretic phenomena but become apparent during
each flight test. Already without visual aids twisting of
load and parafoil is obvious, in particular when performing
highly dynamic maneuvers. No other aircraft offers this kind
of relative motion, so a deeper study seems appropriate.
By a video system further motions get visible, e.g. lateral
displacements. All possible motions are resumed in Figs.
25, taken from [9]. Of course combinations of all these
possibilities can occur as well.
Possible excitations for relative motion are forces that
result from pulling the control-lines and aerodynamic forces,
coming from maneuvers or gusts.

Fig. 5. Relative pitching.

4. Measurement by video
Fig. 6. View of relative motion.

To get a deeper insight in the relative motion between


parafoil and payload-capsule, a sensor-system is required,
which is able to deliver quantitative (numerical) data. Owing
to various difficulties, no conventional sensors can be used.
Therefore a video camera is applied, allowing contactless

measurements from a distance. The camera is fixed to the


load with view to the canopy. Fig. 6 shows a typical sight.
Relative motion is represented as a displacement of the
canopy within the scope of the image [10].

482

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

To extract quantitative measurements from the video data,


the image sequence has to be digitized for further processing
on a PC. Core of the video analysis are so called tracking
algorithms which detect and measure motion in a video
sequence. Tracking makes available positions of significant
points for each video-frame automatically throughout the
entire video sequence. When tracking two ore more points
lying on one single object, information on orientation can be
gained, too.
The canopy is tracked using 4 points, two redundant pairs
of points, providing information on longitudinal and lateral
displacements as well as rotation. The tracking task is performed by a commercial software called WINANALYZE,
developed by the company MIKROMAK [7]. This software
package offers powerful algorithms with a convenient user
interface. The postprocessing of the tracking results is done
using MATLAB. The results are time series with positions,
velocities or e.g. angles for the canopy relative to the load.
The video measurements of a typical flight are shown in
Fig. 7.
XM longitudinal displacement of canopy related to
load,
YM lateral displacement of canopy related to load,
rel twisting of canopy related to load.
Considering also measurements coming from other flight
tests, following results can be stated:

The longitudinal component of the relative motion


hardly exceeds 0.3 m. In frequency domain a 1.1 Hz
oscillation can be found.
The lateral displacement amounts up to 0.5 m and more
in peaks. The frequencies spread from 0.5 to 1.5 Hz.
Twisting reaches 20 ; an oscillation of 5 is always
present. With 0.3 Hz this motion is relatively slow.

5. Models for relative motion


Complementary to the measurement, models of the relative motion have to be derived for further analysis. These
models should feature the significant factors that are responsible for the characteristics of the relative motion, i.e.
multi-body-aspects and aerodynamics. In the following, a
simulation environment for both parts will be designed, a
model of aerodynamics (flight-mechanics), which specifies
the forces that act on the parafoil-load-system, and a multibody-model, which describes how the system reacts on those
forces.

6. Multi-body models
Video analysis yielded data that represent certain system
qualities. A model of this system has to perform preferably
the same way. Thus high demands are made against the
model: it must map the real geometry very closely. All

Fig. 7. Measurements of relative motion.

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

483

elements involved in the relative motion should also be


considered in kinematic modeling [12].
On the other hand for limitation of complexity some
simplifications are required:

Combination of functionally similar bodies.


Elimination of unnecessary degrees of freedom.
Simplification of joints.
Modeling restricted to rigid bodies.

The latter point needs some explanation, because it seems


oversimplified to treat textile elements without any stiff
components as rigid.
6.1. Assumption of rigid bodies
Harness, belts and lines are of a very flexible nature.
Forces can only be transmitted in one single direction;
transmission of moments is not possible. Depending upon
material, elements elongate distinctly under stress, showing
highly nonlinear characteristics. Nevertheless here lines and
belts are to be modeled as rigid bodies. This assumption is
justified by following reasons:
Many possible motions are restricted or prevented by
combining single lines or belts to functional units
(trusses).
Used materials offer practically no lengthening when
stressed; their modulus of elasticity is of the size of steel.
Lift of the parafoil and weight of the capsule produce an
initial tension in harness and suspension lines which is
increased by maneuvering loads; sagging is not to fear.
Mechanical constraints additionally brace the connections between canopy and capsule.
6.2. Couplings
In addition to the rigid bodies there are some other
basic elements of a multi-body-model, generally referred as
couplings. Where individual bodies interact kinematically,
couplings have to be defined. For the modeling of the
parafoil-load-system, following types have been used:
6.2.1. Joints
Usually the participating bodies are connected by joints.
These feature degrees of freedom like observed in the real
system.
6.2.2. Spring-damper-elements
Where its not possible to treat lines, belts or other fabrics
as rigid, flexibility is modeled by the use of springs and
dampers. Here this type has been used just for experimental
purposes at the coupling of canopy and suspension-lines and
at the slider (Fig. 8). That kind of connection is massless.

Fig. 8. Main ALEX-elements.

6.2.3. Constraints
Some interactions between bodies can not be described
appropriately by above-mentioned elements. There, a formulation of a constraint may help. For example the shoulder
strap of the harness is modeled by the condition constant
distance of two specific points.
Using all the assumptions and simplifications described
in this chapter, the kinematic model has been established. In
Fig. 8 the main elements and functional units can be seen.

7. Modeling in SIMPACK
Now from the model a mathematical description of its
characteristics has to be derived. For complex systems like
the one discussed a multi-body simulation package should
be used to avoid oversights and to speed up the modeling
cycle. Here a tool called SIMPACK has been applied,
distributed by the company INTEC [3].
Implementation of the kinematic model in SIMPACK
is accomplished by an interface similar to CAD-programs.
Every single body has to be defined by position, geometry,
mass, and moments of inertia. Tools for the determination of
the static equilibrium and for checking compatibility assist
interconnecting the bodies.
Table 2 shows the elements of the resulting SIMPACK
model, also represented in Fig. 9. In Table 3 the different
couplings are listed that are used to reproduce the relative
motion.
The multi-body-model alone can be used for an analysis
of eigenvalues, natural modes and so on, but is not appropriate to describe the phenomenon of relative motion in general.
So far the model of the relative motion is purely geometrical

484

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

For analysis of the other parts of relative motion the methods


basically are the same, just the models have to be replaced.
The flight mechanical model established in SIMULINK
is a highly simplified representation of lateral motion.
Its purpose is not to map all conceivable effects of a
parafoil-load-system but to provide useful aerodynamic
forces and moments. Equations of motion have been derived
by splitting the 6-DOF model of a rigid parafoil-load-system
into lateral and longitudinal components, neglecting minor
and coupling terms. Finally the mathematical model can be
expressed by [12]:
Ix = L + zp (Ya WTR + F B sin ),
Iz r = N,
m VTR ra = Ya + m g sin ,
ra = p sin TR + r cos TR ,
Fig. 9. SIMPACK model.
Table 2
Model of rel. motion, SIMPACK substitutes
Body

Element

Parafoil
Suspension lines,
left
Suspension lines,
right
Riser, left
Riser, right
Harness, left
Harness, right
Load (capsule)

homog. cuboid
triangular prism,
thickness 0
triangular prism,
thickness 0
rod
rod
rod
rod
homog. cuboid

Mass [kg]
13.1
0.5

zp distance center of gravity-parafoil,


Ya aerodynamic lateral force,
WTR drag at trim state,
F B buoyancy force,
VTR aerodynamic velocity at trim state,
ra aerodynamic rate r,
TR angle of attack at trim state.

0.5
0.2
0.2
0.4
0.4
100

Table 3
Model of rel. motion, SIMPACK couplings
Coupling

Modeled by:

Parafoil-suspension lines (left)


Parafoil-suspension lines (right)
Suspension lines-riser
Riser-harness
Harness-load
Harness-harness (shoulder strap)
Suspension lines-suspension lines (slider)
Load-canopy (control lines)

hinge x-axis, 1DOF


constraint (user defined)
spherical joint, 3DOF
universal x, y joint, 2DOF
hinge x-axis, 1DOF
constraint constant distance
constraint constant distance
force element, point to point

(kinematic). Forces and moments resulting from aerodynamics have been neglected although they definitely affect the
relative motion and thus have to be taken into consideration.
This can be done by an additional aerodynamical (flight mechanical) model. Instead of programming aerodynamic force
laws with SIMPACK, the model for aerodynamics/flight mechanics is established under MATLAB-SIMULINK, which
is more simple and convenient.
From now on this analysis exemplary concentrates on the
relative yawing and the lateral motion respectively. This kind
of relative motion is the most interesting and also complex
one, with the strongest influence on the systems qualities.

Other terms following the international standard nomenclature.

8. Co-simulation
To establish an overall model of relative motion both,
aerodynamics and multi-body mechanics have to be fused.
For each time step of simulation forces and geometric replacements must be computed simultaneously. SIMPACK
provides an interface for communication with SIMULINK
called COSI (Co-Simulation Interface). That means, that
both simulation tools run parallel, exchanging just their results [3]. Each tool uses its own models, data, and solver
algorithms, which guarantees accurate results and comfortable operation. This kind of coupling provides aerodynamic
excitations for the relative motion. Practically COSI is just a
SIMULINK block, which imports many of the capabilities
of SIMPACK to the MATLAB environment.

9. Parameter estimation
With the established model structure it should be possible
to reproduce the measured (video) data as a first application
of the simulation environment. For a best possible match
of simulation and measurement data the parameters of the
model have to be fine-tuned. At the Institute of Flight
Research of DLR Braunschweig, a package for automated
parameter estimation has been developed [11]. This tool

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

485

Fig. 10. Integrated simulation environment.

called FITLAB allows identification of both, linear and


nonlinear systems very conveniently when there already
exists a MATLAB or SIMULINK model.
FITLAB implements maximum likelihood methods. At
a given set of parameters and a certain model structure,
the probability of the occurrence of the measured data is
computed. In an iterative process, parameters are changed
until likelihood reaches its maximum. Hence, simulation is
adapted to give the parameters the most plausible values in
respect to the measured data.
FITLAB also is implemented under the MATLAB environment, so it is possible to complete the model of relative
motion by a module for automated parameter estimation.
The whole system is shown in Fig. 10.
The adjustment of the parameters is based upon a
comparison of simulated and measured data. So a first step
must be to preprocess the measurements. It is of importance
to make the whole set of data consistent, offering no
discrepancies. This can be achieved by the so called flightpath reconstruction.
After that, data is separated into excitations and reactions
of the system. Excitations here are mainly control inputs
(pulling down control lines). Systems reaction can be
observed in changing states, like rates, accelerations, yawing
angle, and relative motion.
Following the principles of system identification [2] the
model is treated with the same excitations as measured
during flight test in the real system. Additionally an initial set of parameters is transferred from FITLAB to the
SIMULINK/SIMPACK-model. For example a total of 13 parameters is responsible for the lateral motion. These initial

values result either from previous knowledge or from analytical considerations. Here parameters have been taken from
both, literature and flight mechanical estimations [1,6,8].
Then the actual simulation starts; SIMULINK calculates
aerodynamic forces and moments, hands them over to
SIMPACK and receives the kinematic states of the parafoilload-system in return. Simulated states are converted in
equivalent measured data for comparison with the real
measurements. FITLAB receives this data and rates the
quality of the simulation. If necessary, parameters are varied
and simulation is started again.
The outlined sequence runs fully automatically, until
the match of simulated and measured data gets best or an
other criterion is achieved. Fig. 11 shows a comparison of
measured (flight test) data and the analogous simulation with
the tuned parameters.
The first graph shows the main excitations, the control
inputs of the left and right actuator. The lateral relative
motion and relative yawing is given for the same time
segment. Simulation and measurement show very similar
characteristics, the model performs well. Dynamics and
amplitudes of the models reaction to excitations match the
flight test data in many respects. Minor differences can be
explained by turbulences and gusts, which are no part of the
model yet.

10. Analysis and simulation studies


The model that is available now is based on real measurements and hence is very credible for certain fields of appli-

486

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

Fig. 11. Lateral motion.

cation. Simulation offers the possibility to answer questions,


which cant be asked in a flight test. For instance it is not
possible to switch discrete forces and moments off for testing purposes. In a simulation this can be done, and so distinctive effects can be studied.
For simulation, a standard excitation has to be defined.
This excitation can be a peak, ramp or step input, but
here an idealized actuator input has been considered more
realistic and meaningful. The main characteristics of a
typical actuator input have been derived from a set of
measured control inputs taken from a test flight. Fig. 12 (top)
shows this artificial input.
10.1. Factors of lateral motion
All physical phenomena of relative motion can be found
in an appropriate branch of the simulation model. So it
is easy to introduce switches to cut away one particular
branch and the respective effect. This provides a powerful

tool for system analysis: all influence factors of relative


motion can be studied separately [10].
For simulation purposes, FITLAB is deactivated, parameters now are not changed any more. The model is charged
with the idealized input and the systems response is calculated for every effect. The simulation results for the major
influences are shown in Fig. 12.
The top graph shows the asymmetric control input for a
left-turn. The three other graphs display the relative yawing
due to one of the three major forces and moments. For
comparison every graph shows the relative yawing with
all influences taken into consideration (all contributions
summarized).
The actuator force acting in the control lines is the main
influence for relative yawing, according to the simulation.
By this force, the load rotates against the parafoil in a very
straight way, because this motion is not aerodynamically but
geometrically driven. But the aerodynamic effects are not
negligible. In particular the yawing moment of the parafoil

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

487

Fig. 12. Simulation study.

shapes the overall relative motion with its characteristic


peaks at the beginning and the end of the control input.
Other simulation studies have been conducted not by
switching branches (forces, moments) of the model off, but
by switching single parameters. This of course has to
regard physical restrictions, some parameters must not be
set to zero. Instead those have been doubled so that their
weighting concerning the other parameters increases. The
differences to normal simulation mode, without changing
parameters, indicate the impacts of those derivatives. So
with every simulation the knowledge of the relative motion
increases.
10.2. Impacts of relative motion
Finally some simulations have been made to study the
impacts of relative motion concerning the overall flight
mechanics. Physically relative motion results in:
increased drag due to an additional local angle of
sideslip and a deformation of the parafoil;
reduction of control efficiency due to increased control
line friction and evasion tendencies of the payload.

shows up, that a parafoil-load-system with relative motion


allowed is better damped and flies more steadily than the
stiff system. Parafoil and load are not so tightly coupled
in the flexible system, both can react on their own on excitations, not affecting the other parts.

11. Conclusion
The phenomena of relative motion in a parafoil-loadsystem have been determined. Because there exists a lack
of measurement data in this field, a new contactless sensor
system using video images has been established. Using this
sensor various flight tests have been conducted, which form
a sound data basis for future analysis. Also the sensor system
should be applicable to other parafoil problems. Based on
the test data a flight mechanical simulation model has been
derived, following the method of system identification. This
model incorporates aerodynamics as well as multi-body
aspects, to get a close match with the experimental vehicle
ALEX. Using this simulation environment relative motion
has been investigated concerning various problems.
11.1. Outlook

For deeper insights the simulation model has been modified to prevent relative motion. The characteristics of this
rigid model can be compared to the flexible model using the
simulation environment and the standard actuator input. It

Parafoil-load-systems vary a lot in configuration and


geometry. So the results of this study can not be applied
to other systems, at least the multi-body-model has to be

488

G. Strickert / Aerospace Science and Technology 8 (2004) 479488

reworked due to other geometry and other terms of inertia.


Since the simulation environment is designed very modular,
changes can be made easily, as well as enhancements and
refinements.

References
[1] W. Gockel, Computer based modeling and analysis of a parafoil-load
vehicle, 14th Aerodynamic Decelerator Systems Seminar, 1997.
[2] P.G. Hamel, R.V. Jategaonkar, Evolution of flight vehicle system
identification, AIAA J. Aircraft 33 (1) (1996) 1028.
[3] Intec, Simpack, Release 7.1, Oberpfaffenhofen, 1998.
[4] T. Jann, Aerodynamic model identification and GNC design for
parafoil-load-system ALEX, 16th Aerodynamic Decelerator Systems
Conference, AIAA 2001-2015.
[5] T. Jann, K.-F. Doherr, Parafoil test vehicle ALEX further development and flight test results, 15th Aerodynamic Decelerator Systems
Conference, AIAA 99-1751.

[6] J. Lingard, Precision aerial delivery/ram-air parachute design, 13th


AIAA ADS Conference and Seminar, Clearwater Beach, 1995.
[7] Mikromak GmbH: Winanalyze, Automatic Motion Analysis, Release
1.4, Erlangen, 1998.
[8] H. Schlichting, E. Truckenbrodt, Aerodynamik des Flugzeuges,
Band 2, Springer, Berlin, 1960.
[9] G. Strickert, T., Jann, Determination of the relative motion between
canopy and load, 15th Aerodynamic Decelerator Systems Conference,
AIAA 99-1754.
[10] G. Strickert, Untersuchung der Relativbewegung fr GleitfallschirmLast-Systeme, Doctoral Thesis, Technische Universitt Braunschweig,
2002.
[11] S. Wei, FITLAB-parameter estimation of nonlinear systems using
Matlab, IB 111-99/32, Institute of Flight Research, DLR Braunschweig, 1999.
[12] L. Witte, Modellierung der Relativbewegung in einem GleitschirmLast-System, IB 111-2000/43, Institute of Flight Research, DLR
Braunschweig, 2000.

You might also like