Conference PaperPDF Available
UAV Guidance with Robotic Total Station for
Architectural Fabrication Processes
Artyom Maxim | Otto Lerke | Marshall Prado | Moritz Dörstelmann |
Achim Menges | Volker Schwieger
1 Introduction
ere is a dramatic change in the conception of architectural production. De-
veloping technologies allow for a higher degree of structural and material ef-
ciency in building construction. rough the integration of novel fabrication
techniques, more emphasis is placed on dierentiated material arrangements
and tailored structural performance. Lightweight, high performance materials
for architectural fabrication empower this change of production and allow for an
increase of functional integration and design possibilities.
Advances in computational design and robotic fabrication techniques have
enabled more automation in the building process, which has recently increased
in importance on construction sites (Meyer 2003). e developments in auto-
mation yield benets regarding the reduction of expenses and the increase of
eciency (Heikkilä and Jaakkola 2003, Gläser etal. 2008). Robotic fabrication
provides more control and adaptability during construction, which has led to the
development of novel production processes.
Lightweight, high performance architectural systems are no longer con-
strained by traditional construction processes, which rely on assemblages of
parts. ese systems can be made onsite at full scale with continuous material
application, eliminating the need for joints that can weaken the otherwise e-
cient structural system. is requires an equally advanced building system capa-
ble of such possibilities as integrated material constraints, fabrication limitations
and structural performance. is paper will discuss some of the developments
required for an aerial fabrication system of a lightweight ber composite struc-
ture. e focus of the paper is the guidance and control of a UAV for ber wind-
ing, utilizing a robotic total station. is enables the fabrication of long span
brous structures in an onsite construction environment.
Previous research at the University of Stuttgart’s Institute for Computation-
al Design (ICD) and Institute for Building Structures and Structural Design
(ITKE) has focused on developing novel fabrication techniques for ber com-
posite building construction. In the ICD/ITKE Research Pavilion 2012 a coreless
winding technique was developed for an onsite fabrication setup (Reichert etal.
2014). An external turntable was used in cooperation with the six axis industrial
145Band 86/2017 p DVWSCHRIFTENREIHE
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
robot to expand the working space of the robotic arm. rough this on-site set-
up, a large scale, continuous ber composite shell was fabricated. e ICD/ITKE
Research Pavilion 2014-15 maximized the reach of the robot by fabricating a
composite shell, in situ, from the inside under low tension (Schieber etal. 2015).
In both examples the highly performative material system was built up incre-
mentally by laying continuous lightweight ber rovings. e interaction of the
bers and the self-forming nature of the composite surface allows for high tol-
erances in production processes. e high strength and precision of the robotic
arms were useful where the tension of the ber and accuracy of winding needs
to be controlled but their limited range of motion and scale constrained the ma-
terial system to relatively small scale architectural demonstrators.
Aerial vehicles provide one possible solution to the limitations of industri-
al robotics in architectural production of ber composite structures. ey are
untethered and have a less constrained range of motion. Aerial vehicles have
already been used in the construction industry for lming and surveying tasks
(Burger etal. 2016, Rose 2016), but have had limited use in construction pro-
cesses for material and environmental interface. ere is a growing eort to
change this through developing architectural research. Several projective pro-
jects at the ETH Zurich (Mirjan etal. 2013), University of Pennsylvania (Maho-
ny etal. 2012) and the Architectural Association in London have utilized aerial
vehicles in novel construction systems. ey have shown that aerial vehicles can
pick, place and pull objects having a limited payload within a controlled pro-
duction environment. Materials such as rope have been used to fabricate tensile
structures, showcasing the potential use of aerial vehicles for ber composite
construction. ese examples demonstrate the inherent exibility and agility of
micro aerial vehicles in interacting with environment and set course towards the
new, exciting applications in architectural design and construction. Although
these examples made quite remarkable developments in UAV navigation within
a controlled fabrication environment and showed the interaction with objects
and materials, it is still to be demonstrated how these processes can transition
from the controlled lab-like environment to a geometrically complex, onsite or
in situ construction environment.
In order to transition from the controlled lab environment to the onsite fab-
rication setup, dierent localization strategies may be required. e positioning
systems employed for the afore mentioned for the control of the aerial vehicles
can be unsuitable for onsite construction environments. Motion capture setups,
enhanced GPS tracking (RTK-GNSS) and onboard SLAM algorithms, common-
ly used setups for localization of aerial vehicles, are useful in some construction
environments but may encounter problems with signal shadowing, reectance,
magnetic interference, light pollution or feature recognition when applied to
onsite construction.
146 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
is paper provides an alternative method for aerial vehicle localization and
control utilizing standard construction equipment suitable for onsite fabrication.
Robotic total stations are portable, robust machines capable of accurate position
estimation in various indoor and outdoor conditions over large distances. ey
are useful for site survey-
ing, digital scanning and
motion tracking. Heavy
machinery such as grad-
ers and pavers have used
robotic total stations for
the localization and con-
trol on construction sites
(Fig.1 and Fig.2).
Robotic total stations
have been used for track-
ing of aerial vehicles (Fitz
2013), though guidance
of UAVs by tachymetry
is not common. Never-
theless tachymetric guidance of UAVs may become more prominent if demands
on accuracy rise and operational scenarios become complex, as with onsite con-
struction of ber composite structures. Furthermore this positioning method is
reliable for operation even in challenging environmental, weather and lighting
conditions, where other methods might not be suitable. e given architectur-
al fabrication task requires an optimal performance during ight operations in
exible environments. Regarding the technical implementation of tachymeter
guidance, the existing machine guidance methods used for construction equip-
ment can be transferred to UAV applications with slight modications, which
will be shown in the following chapters.
Fig. 1:
3D Control, Exam-
ple of a Mainline
Stempfhuber 2008
Fig. 2: Road Constructions (Certus Verlag AG, Contribution
by Terradata AG, last access on 13.09.2016)
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
2 UAV – Guidance System
2.1 Description of the UAV
A custom aerial vehicle was built using both o the shelf electronic and propul-
sion components and custom made frame parts. e brushless motors, motor
controllers, battery and propellers, used for the propulsion system, are standard
for self-built quadrotors. e car-
bon bre plates and aluminium
proles were used for the frame
(Fig.3). Electronic setup includes
an autopilot board, onboard com-
puter, several radio and power
modules, USB-serial converters.
e autonomous vehicle op-
erates using two main hardware
components – the autopilot and
the onboard computer. A “Pix-
hawk” autopilot platform, with
“PX4” rmware (Fig.4), was cho-
sen as the core of the system, as
it has a growing developer community, a substantial online documentation and
has a completely open source hardware and soware system. e Pixhawk is
an embedded computer system with integrated MEMS based inertial, magnetic
and barometric sensors and hardware interfaces for controlling an autonomous
aerial vehicle. e PX4 is “a node-based multithreaded open source robotics
framework for deeply embedded platforms” (Meier etal. 2015). Pixhawk is used
in conjunction with the onboard computer “Odroid XU4”, which is an ARM
single board computer that is running Ubuntu, with Robot Operating System
(ROS). ROS is framework for programming and controlling robots, machines
and systems of any kind (ROS 2016). It provides all necessary tools for creating
complex and reliable computerized systems. In this project, ROS is used as a
high level control mechanism for the vehicle. It is able to run navigation algo-
rithms, control packages with advanced logic and external interfaces and serves
as a bridge to the autopilot.
e power system includes a 5Ah battery, two power modules, a 5V5A power
supply for onboard computing and a Pixhawk power module, which not only
provides power to the autopilot, but also data about the current voltage and cur-
rent state of the battery. An onboard WiFi adapter is also included for ground
computer connection, accessing the terminal of the onboard computer, upload-
ing code and ight programs and observing autopilot state. Additionally two
Fig. 3: UAV
148 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
radio modules – FrSky radio receiver for manual remote control and Laird
RM024 2.4 Ghz with ½ wave 2dBi antenna are installed. A pair of RM024 radio
modules were used to robustly transfer localization data from the ground com-
puter, connected to the total station, to the onboard system of the aerial vehicle.
e ight range is only limited by the radio receivers. e operating distance
is around 1 km. e ight time on one battery charge is 10to 15minutes. e
payload is 500grams.
2.2 Guidance with Robot Total Station
In this work a Leica MS60 robotic total station has been used as a positioning
sensor. MS60 is a state of the art robotic total station. It provides position accu-
racies of 3to 5 mm in kinematic mode (Leica 2016) and thus complies with high
accuracy demands. e MS60 can be dened as a multi-sensor system. Besides
including several sensors to measure temperature, pressure, humidity, the to-
tal station also includes relevant sensors for position determination, horizontal
and vertical angle measurement, as
well as electronic distance measure-
ment, the EDM. e most impor-
tant features for this investigation
is the Automatic Target Recogni-
tion (ATR) sensor which allows to
recognize and track the target. For
each measurement the station pro-
vides X, Y, Z-coordinates. In this
investigation the total station has
been used in combination with a
360° prism Leica GRZ101 (Fig.5).
e instrument provides several
useful tools for kinematic tracking
Fig. 4:
Auto pilot
(left), Odroid
XU4 (right)
Fig. 5: Multi Station Leica MS60 (left), Leica
360° prism GRZ 101 (right)
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
including: Dynamic Lock and Power Search functions (Leica 2016), good target
holding quality, high tracking rate of 20 Hz, low rate of data gaps, tolerance to
target occlusion and an enhanced target prediction phase. e limitations are the
tracking distance, the general problem of retention of the line of sight, valid for
all tachymeter applications and unfavourable atmospheric conditions.
2.2.1 Transfering Estimated Positions to the PX4 Autopilot
e guidance methods are implemented using the robot operating system (ROS)
platform programming framework. e system consists of several ROS nodes,
running on both, ground and onboard computers and connected by a radio link.
e “Tracking” node communicates with the tachymeter using a serial connec-
tion to receive the X,Y, Z-coordinates of the measured prism position, mounted
on the vehicle. e implementation of this node is based on an existing open
source Python library for communicating with Leica total station instruments
using the “GeoCom, an ASCII based request-response protocol, which makes
it compatible with many computer platforms, e. g. microcontroller, smartphone
or a PC. Aer receiving coordinates from the “Tracking” node, a “Sender” node,
passes the data via serial connection to the RM024 radio module, congured
as a “Server” to transmit position information to the onboard computer of the
UAV. e data gets received by the “Client” RM024 module, read by the ROS
“Receiver” node and shared for other onboard ROS system nodes, for use. e
coordinates can then be accessed by a main ROS “Control” node, which handles
safety checks for data consistency and controls the execution of custom ight
instructions. e “Control” node passes measured real-time position and com-
mands to the “Mavros” node – a ROS package, which provides simple means of
communication and control of the aerial vehicle. Using Mavros, the “Control
node can obtain and utilize any data from the autopilot, including its current at-
titude, state, mode, battery charge level and multiple others. Mavros then passes
commands to PX4 autopilot over a serial connection, which in turn controls the
ight behaviour of the vehicle. Additionally, the “Rosbridge” node on the ground
computer passes measured coordinates to a visual scripting interface, integrated
with a common architectural CAD soware, which is able to interactively dis-
play, store and evaluate trajectories of the ight path (Maxim 2016). e main
function of this soware within this project is the real-time monitoring of ight
operations on the ground computer.
2.2.2 Integration of Position Data into Autopilot
PX4 Autopilot is able to use externally measured position data for localization, in
relation to its initial origin, which resets to zero at startup. In the developed set-
150 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
up, a local position estimator (LPE) soware module of autopilot is used, which
is responsible for processing the position data from internal or external sensors.
LPE is based on an extended Kalman lter algorithm, oen used for nonlinear
discrete systems. e Kalman lter is a set of mathematical equations that pro-
vides an ecient computational (recursive) means to estimate the state of a pro-
cess, in a way that minimizes the mean of the squared error. e lter is powerful
in several aspects: it supports estimations of past, present, and even future states
by prediction, and it can do so even when the precise nature of the modeled
system is unknown (Welch and Bishop 2004). LPE is capable of fusing multiple
position sources, such as GPS, Motion Capture or Computer Vision, potentially
improving the precision of the resulting estimate, however, in the presented set-
up, only the robotic MS60 Multi Station has been used, as it alone provides posi-
tion estimates for moving targets with millimeter precision. From LPE estimator,
the position data gets passed on to another soware module of PX4 autopilot,
the “Position Controller”. It is based on a proportional-integral-derivative (PID)
controller method and is responsible for calculating the amount and direction
of force that the vehicle needs to produce in order to reach the destination tar-
get setpoint. e position controller has multiple, separate closed-loop systems,
controlled by PID controllers, for dierent aspects of the motion of the vehicle,
e. g. vertical and horizontal movements as well as velocities. ese controllers
need to be tuned properly, for achieving acceptable motion accuracy of the ve-
hicle. If the controllers are untuned, the vehicle tends to oscillate around a target
setpoint, never actually reaching it. In worst cases, the amplitude of oscillations
will increase and the vehicle becomes completely unstable. In this project, the
PID controllers were tuned using an heuristic approach, which has proven to be
sucient for eliminating oscillations. Detailed description on the PID controller
theory and tuning can be found in Mann etal. (2005) or Busch (2012). e out-
put from the position controller is passed to the “Attitude Controller, a soware
module, which gets the attitude information (roll, pitch yaw) from an onboard
IMU and is as well based on a PID controlled closed-loop system. e attitude
information is then passed to the “motor mixer” and the “motor driver” soware
modules, which control the speed of the rotors.
2.2.3 Local Reference Frame Orientation
An important point is achieving correct reference frame orientation of the total
station. is coordinate frame can be established by the “free stationing” tech-
nique. e autopilot utilizes an internal compass sensor, for determining its ori-
entation heading in space. e autopilot navigation frame is North-East-Down
(NED), which means its X, Y axes are pointing along the magnetic North, East
directions respectively and the Z axis points perpendicular to the other axes,
151Band 86/2017 p DVWSCHRIFTENREIHE
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
downwards. By convention, ROS operates in another coordinate frame, the
ENU (East-North-Up). erefore the coordinates form the instrument should
be converted to ENU format. e Mavros node (Fig.6) is responsible for con-
verting ENU coordinates to the NED convention and passing them to the PX4
Autopilot. e information about the vehicle’s orientation heading is provided
by an onboard magnetometer and is the main source of the heading informa-
tion. is sensor, which measures the magnetic eld of the earth to dene the
orientation of the platform, is easily aected by various magnetic disturbances,
caused by onboard power cables, motors and even by large metallic objects in
the surrounding environment. In this case, the onboard compass heading could
be suciently distorted during the ight operation and become unreliable or
imprecise. e resulting eect is a misorientation during the ights causing de-
viations from the desired trajectory. In this context the problem of an insucient
heading determination of the UAV might be present in some cases, but wasn’t
addressed in this investigation, as the performed ights have shown that slight
misalignment of the frames by few degrees is not critical for the given applica-
tion. During the performed experimental ights only small deviations from the
expected trajectory during the vehicle’s movement could be detected. e solu-
tion for this circumstance would be an additional orientation estimation source,
e. g. a computer vision system, use of the reference trajectory or the use of the
eective trajectory sections according to Beetz (2012).
2.2.4 Design and Execution of Custom Flight Trajectories
e direct control of vehicle behaviour and functionality is achieved using a cus-
tom “Control” node as described previously (Fig.6). is node is a functional
Fig. 6: Guidance System Architecture
152 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
core” of the system, allowing for the execution of ight programs in an auto-
mated mode, stored onboard. It serves as a communication hub as well, feeding
commands and position measurements to the autopilot and reading vehicle sta-
tus and passing the information on to the user. e ight trajectories are pro-
cessed by the visual scripting interface, to generate a custom-formatted code,
consisting of an index, position and orientation information for every point
along the ight path. e ight code generation is tightly connected to the de-
sign workow, where the ight paths can either be drawn manually in a CAD
soware, or produced automatically from the computationally dened lament
winding order. is ight program is then sent and stored in the onboard “Con-
trol” node and is later executed line-by-line during the ight. At startup there is
a discrepancy between the internal local position, dened as zero by autopilot
rmware, and its real-world position, dened by the measurement in the coor-
dinate frame of the total station. erefore, one of the “Control” node tasks is to
oset both, measured position and trajectory targets, by the initial startup world
position oset, which allows to coordinate everything to the same superordinate
reference frame. e oset must be considered for a 3 dimensional case. e
target setpoints, the measured prism position, as well as the architectural fab-
rication model inside CAD soware can now be represented in the ENU world
frame, making the system easy to use by avoiding unnecessary frame rotations.
3 Test Results
e guidance system was tested in context of a full-scale lament winding exper-
iment. It is based on a freestanding conguration of four winding frames, with
one 4 m high in the center and three 1.5 m high, equally spaced around, to form
a triangle with a side of 7 m (Fig.7). Aer arranging the frames, the MS60 Total
Station was used to scan the environment to nd both, the surrounding bound-
aries of the fabrication space and the precise position of the frames themselves.
is allowed the construction of the ight trajectories around measured objects
in the CAD soware. e resulting ight path is then converted into custom-for-
matted instructions for the vehicle and uploaded to the onboard system. Aer
the total station guidance has been set up and the vehicle was receiving its meas-
ured position, the user activated the automatic ight mode for the quadrotor to
start executing the program.
Fig.8 exemplarily depicts the eective ight path and the reference ight path
between the waypoints.
In order to evaluate the control quality for the tachymetric guided UAV, the
3Ddistances between reference and eective ight path have been analysed. is
approach was inspired by 2Devaluation approaches, e. g. represented in Lerke
153Band 86/2017 p DVWSCHRIFTENREIHE
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
und Schwieger (2015) or Beetz (2012), but has been adapted for a 3Dscenario.
e adapted method analyses the shortest distance between the measured posi-
tions at timestampi and the reference ight path. To determine this distance a
cross-product approach has been used to calculate the plumb line through the
measured position perpendicular to the reference ight path. Subsequently the
length of the plumb line has been calculated and analysed. Details on cross-prod-
uct theory can be found at e. g. Merziger and Wirth (2010).
In order to examine the control quality, minimum and maximum distances,
as well as the mean value and the root mean square (RMS) have been calculated.
e formula for the RMS is given below:
Fig. 7: Experiment Site and Construction Object
Fig. 8: 3-D Flight Path (shown exemplarily for one round)
154 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
iii ii ii
di – lenght of the plumb linei, nnumber of measurements,
– coordinate triple for pointi of the reference trajectory,
– coordinate triple for pointi of the effective
Tab. 1 shows the summarized results on the above described evaluation method.
e histogram in Fig.9 shows the distance sample distribution for the com-
plete recorded ight path. e y-axis represents the amount of samples, the
x-axis represents the calculated distances. e histogram reveals that over 40 %
of the sampled distances are below 100 mm. Around 25 % of the samples are
between 100 mm and 200 mm and the remaining 35 % are between 200 mm and
1,000 mm. e results show a wide value range and therefore a wide range of the
resulted control quality.
Tab.1: 3D Flight Path Evaluation
Section of the Flight
Min. Distance
Max. Distance
1 4 967 273 371
2 6 919 195 284
3 3 906 262 352
Complete (1 + 2 + 3) 3 967 235 329
Fig. 9: Distribution of 3Ddistances to reference trajectory for complete ight path;
number on bars equal to number of samples
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
To take a closer look
on the situation, the ight
movement has been sepa-
rated into a horizontal and a
vertical component. Firstly,
an evaluation of the hori-
zontally projected trajectory
has been conducted. Fig.10
depicts a ight path section
with the reference ight
path, the eective ight path
and the plumb foot points,
projected on the x-yplane.
e numerical values for
the horizontal component of
the movement are summarized in Tab.2. e values show, that the lateral devia-
tion of the eective ight path is in the range between 0 mm and 300 mm. us
the guidance quality, referred to the x-yplane, is approximately three times better
than for the 3Dcase. e histogram in Fig.11 depicts the samples distribution
Tab. 2 : 2D ight path evaluation (referenced to x-y plane)
Section of the Flight
Min. Distance
Max. Distance
1 1 303 96 129
2 <1 296 80 99
3 <1 186 65 82
Complete (1 + 2 + 3) <1 303 81 104
Fig. 10: 2-D Flight Path, projected to x-y plane
Fig. 11: Distribution of 2Ddistances to reference trajectory, projected to x-yplane;
number on bars equal to number of samples
156 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
of the plane based method. e histogram shows that approximately 70 % of the
overall distances are below 90 mm.
For the completion of the analysis, the evaluation of the vertically projected
trajectory has been performed. Fig.12 shows the 2-Dight path, referenced to
the y-zplane. It is noticeable from Fig.12, that the eective ight path exhibits
large vertical deviations from the reference ight path. Tab . 3 summarizes the
numerical values for the vertical component of the ight movement.
e analysis shows, that the vertical component of the ight movement has
the greatest impact on the examined distances. us, the altitude control (along
z-axis) has the worst performance. e histogram in Fig.13 shows, that only
40 % of the calculated distances are smaller than 100 mm.
Fig. 12: 2-D Flight Path, projected to y-z plane
Tab. 3 : 2D ight path evaluation (referenced to y-z plane)
Section of the Flight
Min. Distance
Max. Distance
1 0 966 260 361
2 0 916 181 273
3 0 901 251 344
Complete (1 + 2 + 3) 0 966 222 319
Fig. 13: Distribution of 2Ddistances to reference trajectory, projected to y-zplane;
number on bars equal to number of samples
157Band 86/2017 p DVWSCHRIFTENREIHE
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
According to the datasheet of the tachymeter, the accuracy is 3 mm to 5mm
for positioning in kinematic mode and thus its inuence on the resulted guid-
ance accuracy is neglectable. e values from Tab. 1 , Tab.2 and Ta b. 3 comprise
frame misalignment, PID controller parameter settings, as well as ight dynami-
cal aspects and further unknown inuences. Especially the z-component, under-
lying the gravitational forces, has the largest share of the reduction of guidance
4 Conclusion and Outlook
e guidance of UAVs by robotic total stations is possible and enables the ex-
ploration of novel large scale on site automated fabrication processes for archi-
tectural structures. e conducted experiments have shown, that tachymetric
guided UAVs fulll the quality requirements for fabrication of large scale on site
construction on the example of a lament structure.
For the 3Dcontrol quality a RMS of 329 mm could be detected for the com-
plete ight path. e horizontally projected evaluation of the control quality, ref-
erenced to the x-yplane, revealed a RMS of 104 mm and the vertically projected
evaluation, referenced to the y-zplane, indicated a RMS of 319 mm. Although
these results suggest a marginal control quality, they comply with the given re-
quirements for core-less lament winding methods. is method allows for a
relatively large tolerance range, as the accuracy of the fabricated piece is deter-
mined by the winding frame and is not a direct representation of the ight path.
e reciprocal constraints of payload, drone size, battery capacity and ight
time dene boundary conditions for the presented fabrication method. Future
trajectories of development towards a specic fully functional fabrication meth-
od for large scale ber composite parts could include stationary positioning of
the ber spool, thus the spool is not part of the payload. In this scenario the
UAV is mainly utilized as a ber guidance and positioning vehicle. is would
allow for a more continuous fabrication process, as the reduced payload allows
for longer ight time and a ground supported spool could be signicantly larger,
requiring less frequent spool changes. A second aspect relevant for scaling of this
technology is the balance between slack and tension force in the ber roving.
While UAVs allow for low tension stringing of lament over large distances with
a comparatively low level of precision, a robotic arm allows for high precision
positioning of lament with a substantial amount of pre-tension within a limit-
ed operation space. Combination of industrial robots and UAVs render a very
promising future opportunity, as it combines both the UAVs benet of a large
operation space with local robotic precision and strength to pre-tension bers
over long distance.
158 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
Beyond the specic application for large scale fabrication of lament struc-
tures, the general control and localization method using a robotic total station
has wider elds of application in automated construction processes. For the
emerging research eld of mobile fabrication agents, a robust localization sys-
tem can be an important key technology to enable coordination of fabrication
robots oen proposed to be utilized in large quantities with semi-autonomous
group behaviour. Beyond positioning of fabrication machinery, total stations can
fulll multiple tasks, e. g. tracking of the fabrication process, thus allowing to
inform the fabrication agents behaviour as an inline control mechanism, or nal
measuring and quality control.
Sensor integrated fabrication strategies, such as total station enabled aerial ve-
hicle localization and control, set the base for adaptive modes of fabrication that
allow for both, higher accuracy and control in known processes, and exploration
of novel fabrication strategies.
Funded by Volkswagen Foundation.
Augugliaro, F.; Lupashin, S.; Hamer, M.; Male, C.; Hehn, M.; Mueller, M. W.; Willmann,
J. S.; Gramazio, F.; Kohler, M.; DAndrea, R. (2014): e Flight Assembled Architecture
installation: Cooperative construction with ying machines. In: IEEE Control Sys-
tems, vol.34, no.4, pp.46–64, Aug. 2014.
Beetz, A. (2012): Ein modulares Simulationskonzept zur Evaluierung von Positionssen-
soren sowie Filter- und Regelalgorithmen am Beispiel des automatisierten Straßen-
baus. Dissertation, Universität Stuttgart.
Burger, S.; Elein, A.; Völter, U. (2016): Photo-based-scanning als Erweiterung des inge-
nieurgeodätischen Leistungsspektrums. In: DVW e. V. (Hrsg.): UAV 2016 – Ver mes-
sung mit unbemannten Flugsystemen. DVW-Schrienreihe, Band 82/2016, Wiß ner-
Verlag, Augsburg 2016, S.175–189.
Busch, P. (2012): Elementare Regelungstechnik. 8., überarbeitete Auage, Vogel Buchver-
lag, Würzburg.
Fitz, D. (2013): Bestimmung von Referenz-Positionen für die photogrammetrische Aus-
wertung von UAV-Bildügen mittels einer automatisierten Totalstation. Master e-
sis, Institute of Engineering Geodesy, University of Stuttgart, Stuttgart (unpublished).
Gläser, A.; Möhlenbrink, W.; Schwieger, V. (2008): A Contribution to Construction Ma-
chine Automation. Journal of Applied Geodesy2 (2008). DOI: 10.1515/JAG.2008.011.
Heikkilä, R.; Jaakkola, M. (2003): Automatic Control of Road Construction Machinery
– Feasibility and Requirements. 20th International Symposium on Automation and
Robotics in Construction, Eindhoven, Netherlands.
159Band 86/2017 p DVWSCHRIFTENREIHE
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
Leica, 2016:
Lerke, O.; Schwieger, V. (2015): Evaluierung der Regelgüte für tachymetrisch gesteuerte
Fahrzeuge. In: zfv – Zeitschri für Geodäsie, Geoinformation und Landmanagement,
He 4/2015, 140.Jahrgang. DOI: 10.12902/zfv-0080-2015.
Mahony, R.; Kumar, V.; Corke, P. (2012): Multirotor Aerial Vehicles: Modelling, Estima-
tion, and Control of Quadrotor. IEEE Robotics & Automation Magazine 19(3): 20–32.
Mann, H.; Schielgen, H.; Frorier, P. (2005): Einführung in die Regelungstechnik.
10.Au., Carl Hanser Verlag, München Wien.
Maxim, A. (2016): MAV Assisted Fabrication Strategy for Large Scale Fiber Composite
Structures. Master esis, Institute for Computational Design, University of Stuttgart,
Stuttgart (unpublished).
Meier, L.; Honegger, D.; Pollefeys, M. (2015): PX4: A node-based multithreaded open
source robotics framework for deeply embedded platforms. 2015 IEEE International
Conference on Robotics and Automation (ICRA), Seattle, WA, 2015, pp.6235–6240.
DOI: 10.1109/ICRA.2015.7140074.
Merziger, G.; Wirth, T. (2010): Repetitorium Höhere Mathematik. 6.Au., Binomi Verlag
Barsinghausen, ISBN: 978-3-923923-34-2.
Meyer, H. J. (2003): Anwendung von geodätischen Positionsmesssystemen in Straßenbau-
maschinen. Baumaschinentechnik 2003 – Erfahrungen, Methoden, Innovationen –
Fachtagung 20./21.03.2003 Dresden. He Nr.23 der Schrienreihe der Forschungsv-
ereinigung Bau- und Baustomaschinen.
Mirjan, A.; Gramazio, F.; Kohler, M.; Augugliaro, F.; D’Andrea, R. (2013): Architectural
fabrication of tensile structures with ying machines: BT – Green Design, Materials
and Manufacturing Processes, 513–518.
Reichert, S.; Schwinn, T.; La Magna, R.; Waimer, F.; Knippers, J.; Menges, A. (2014): Fi-
brous structures: an integrative approach to design computation, simulation and fab-
rication for lightweight, glass and carbon bre composite structures in architecture
based on biomimetic design principles. Computer-Aided Design, 52, pp.27–39.
ROS (2016):, last access 21.11.2016.
Rose, A. (2016): Einsatz von UAV bei Katastervermessungen – ein Update. In: DVW e. V.
(Hrsg.): UAV 2016 – Vermessung mit unbemannten Flugsystemen. DVW-Schrien-
reihe, Band 82/2016, Wißner-Verlag, Augsburg 2016, S.191–201.
Schieber, G.; Koslowski, V.; Knippers, J.; Dörstelmann, M.; Prado, M.; Vasey, L.; Men-
ges, A. (2015): Integrated Design and Fabrication Strategies for Fibrous Structures. In:
Modelling Behaviour (pp.237–245). Springer International Publishing.
Stempuber, W. (2008): Navigation of Construction and Agriculture Machinery, Swiss
Federal Institute of Technology, ETH Zurich, Switzerland, 2008.
Vidal, E. (1991): New Formulation and Improvements of the Nearest-Neighbour Ap-
proximating and Eliminating Search Algorithm (AESA). Pattern Recognition Letters,
Vol.15, Issue1, (1994).
Welch, G.; Bishop, G. (2004): An Introduction to the Kalman Filter. Technical Report
TR 95-041, Department of Computer Science, University of North Carolina at Chapel
160 DVWSCHRIFTENREIHE p Band 86/2017
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
Willmann, J.; Augugliaro, F.; Cadalbert, T.; D’Andrea, R.; Gramazio, F.; Kohler, M. (2012):
Aerial robotic construction towards a new eld of architectural research. International
journal of architectural computing, 10(3), pp.439–459.
Artyom Maxim | Marshall Prado | Moritz Dörstelmann | Achim Menges
Institute for Computational Design, University of Stuttgart
Keplerstrasse 11, 70174 Stuttgart
Otto Lerke | Volker Schwieger
Institute of Engineering Geodesy, University of Stuttgart
Geschwister-Scholl-Strasse 24D, 70174 Stuttgart
161Band 86/2017 p DVWSCHRIFTENREIHE
Maxim et al., UAV Guidance with Robotic Total Station …
© Wißner-Verlag
... It is lower than what is available while using the prism tracking RTS. Paper [46] provides an alternative method for aerial vehicle localization and control utilising standard construction equipment suitable for onsite fabrication. The three-dimensional (3D) distances between reference and effective flight path have been analysed in order to evaluate the control quality for the RTS guided UAV. ...
... For example, in [50], it was shown that deviations of 10-100 m have to be expected during a GNSS outage of 30 s, when MEMS inertial sensors are used on a UAV. However, in the paper on the control of a UAV with a RTS, the following results are achieved for the 3D control quality-a Root Mean Square Error (RMS) of 329 mm was detected for the complete flight path [46]. The horizontally projected evaluation of the control quality, referenced to the x-y plane, revealed a RMS of 104 mm and the vertically projected evaluation, referenced to the y-z plane, indicated a RMS of 319 mm. ...
... However, the time drift in the horizontal plane has mostly occurred during the strong wind tests and it has resulted from the wind pressure on the device, Table 2. In the paper on control of a UAV with a RTS 3D position, quality has been measured for the complete flight path [46]. In this case, the working space was also limited to a very small space, in a cube of a size of not more than 15 m. ...
Full-text available
This paper describes an experimental test campaign while using an Unmanned Aerial Vehicle (UAV) and measuring the obtained UAV positions during different flight tasks and in different operative conditions. A new test procedure has been presented and tested for different devices in various weather conditions. This paper describes and analyses the measurements of the flight trajectory of the UAV that was performed with the use of a robotic total station (RTS), as compared to the design data and the data recorded in the internal memory of the UAV. Five different test tasks have been conducted. The obtained results have allowed for the assessment of the correctness of task performance as compared to the design and to determine the flying accuracy of the entire UAV set. The proposed set of tasks can be successfully utilised to control the correctness of operation of various types of UAVs and it may be implemented as a universal test to verify the algorithms optimising take-offs and landings, test flights of the objects, as well as flight planning in various terrain and weather conditions, which will increase the safety of the flights while using UAVs.
... 3D models of buildings and structures can be created using drones, but sometimes it is difficult to fly with drones in a limited space. A combination of total station and drone solves this problem [3] [4] [5] [6] [7] [8]. Total station measures the position of the drone and it sends this information to the drone program. ...
Full-text available
Using drones with different purposes than only taking photos is nowadays the main direction of drone development. Drones are made for package delivery, people transport, etc. Drone equipped by GNSS RTK and prism can be used as orientation point for the free station. The idea is using drone to get coordinates of total stations inappropriate for GNSS. such as high buildings and forest. The drone can fly above the obstacle causing inappropriate, so the GNSS will compute the position coordinates correctly. Total station will measure distance and angles on prism to get free station coordinates. This article deals with the accuracy of using two points in the free station task. Accuracy of measurement and data is based on real values. Drone can be used as the target if it is not windy, the position accuracy of the target on drone is 5 cm. Wind has no effect on the vertical position accuracy of the the drone. The results show that the same principles and limitations must be observed when measuring the free station task. Horizontal angle between orientation points must be bigger than 100 gon and the zenith angle must be at least 50 gon. The distance between orientation and free station must longer than consequent measured points.
... Either RTS or radar tracking require line-of-sight access to the UAV, which can be not available during significant portions of the survey depending on the operating environment. Moreover, since the originating use of RTS is the track of stationary or almost stationary targets, tracking rate is low and either latencies (Roberts and Boorer, 2017) or tracking loss (Maxim et al., 2017) can occur. ...
... Any real construction machine describes a similar movement, namely travelling at low speeds and without sudden direction changes. On the other side, if the system is used for kinematic applications that involve irregular movements or quick direction changes, like an operator carrying a reflector on a meas- urement pole or tracking and guiding of an UAV (MAXIM et al., 2017), the reachable ac- curacy and reliability need to be evaluated. The second scenario includes variable speeds and interruption of the line-of-sight during a random walk. ...
Conference Paper
Full-text available
A network comprised of Robotic Total Stations allows the continuous tracking of a moving reflector even if obstructions interfere with the line-of-sight. The following paper presents a real time network system that connects two instruments, the Leica TS30 and TS16, which communicate with a central computer while tracking a moving 360° reflector. These are positioned in the same coordinate frame and track the reflector in a synchronized manner. If the line-of-sight of one instrument is interrupted, the total station switches into a passive state and will continue to ‘‘blindly’’ track the reflector until a new line-of-sight is reestablished. To verify the system’s performance, two different scenarios are shown. In the first case, a reflector is fixed on a calibration rail and is travelling with constant speed. The second case analyzes an irregular movement described by a moving person. In each scenario, obstacles interrupt the line-of-sight in a controlled or random manner. The outcomes put a light on the achieved positioning quality and the tracking process while obstructions occur.
... In our case, the prism is placed on the drone. The feasibility of this system combined with a drone has been proved by a team of researchers from Stuttgart [33]. This system has the advantage of being very accurate (precision of a few mm) and uses very common equipment used on construction sites. ...
The additive manufacturing of real scale structures using UAVs (drones) is a new discipline with challenges as wide as the possibilities it opens up for the future. UAVs must not be seen as the only way of robotizing future construction sites, but in combination with other kinds of robots. This adequate combination is indeed likely to reduce the influence of factors that usually badly affect the quality and profitability of construction projects, such as human factors, execution slowness, insecurity, insufficient communication between the stakeholders, weather conditions, strikes, lack of skilled labor, etc. The aim of this research, carried out jointly by MIT and UCLouvain since 3 years, was to lay the necessary groundwork, still not explored elsewhere, in order to prove the feasibility of building real-scale structures, in particular masonry structures, with big custom-built drones. In particular, the objective was to investigate the drones precision, their behavior while transporting, handling and laying loads, but also to draw the first guidelines for the design of “Drone compatible” construction elements: their shape, the way they should be assembled together, how to minimize their weight, how to connect them together, how to ensure their stability. This publication summarizes the work carried out so far in this field, provides the results of the laboratory tests and proposes development and improvement paths for the future. In particular, lab tests with a big drone assembling different kinds of more and more complex construction elements are commented. Several conclusions can be drawn from the study, the first one being that the research is worth going beyond the step of proving the feasibility. Indeed, it shows that using UAVs for the construction of future real scale structures is certainly not a utopia and is very promising. However, it requires further developments, not only about the drone themselves (guiding systems, handling systems, robustness, power supply), but also about the way to pass from the laboratory stage to the construction of real structures with a complex geometry, composed of slabs, walls, connections and finishing.
... Either RTS or radar tracking require line-of-sight access to the UAV, which can be not available during significant portions of the survey depending on the operating environment. Moreover, since the originating use of RTS is the track of stationary or almost stationary targets, tracking rate is low and either latencies (Roberts and Boorer, 2017) or tracking loss (Maxim et al., 2017) can occur. ...
Micro-Unmanned Aerial Vehicles (UAVs) are flexible observation platforms suitable to cover inaccessible areas on demand. Accordingly, huge attention is deserved towards development of miniaturized sensing technologies compliant with UAV payload constrains and capable of providing high-resolution images of the region under test. As a contribution to this topic, the paper presents a prototype UAV system for radar imaging made by a commercial micro-UAV equipped with a miniaturized low power radar, whose imaging capabilities are enhanced by a properly designed data processing strategy. Such a strategy involves a processing step performed in time domain, which accounts for a procedure devoted to compensate flight altitude variation and a Singular Value Decomposition (SVD) based noise filtering approach. After, the focused images of the surveyed scenario are obtained by using a microwave tomographic approach, which integrates data about UAV position and faces the imaging as a linear inverse scattering problem. A feasibility experiment, carried out to test the operational mode of the assembled system, is presented. The obtained results corroborate that the integration of GPS and flight altitude information into the microwave tomography approach allows valuable radar imaging capabilities in terms of target localization accuracy.
... Either RTS or radar tracking require line-of-sight access to the UAV, which can be not available during significant portions of the survey depending on the operating environment. Moreover, since the originating use of RTS is the track of stationary or almost stationary targets, tracking rate is low and either latencies (Roberts and Boorer, 2017) or tracking loss (Maxim et al., 2017) can occur. ...
Full-text available
The aim of the research presented in this paper is to examine the potential of integrated computational design methods for the development of complex fibre-based structures for architectural applications. While in today's building industry structural systems still rely on a set of well known typologies, integrated computational design strategies offer the opportunity to break down traditional strategies and develop new resource efficient architectural systems by synthesizing the material constraints, structural performance and digital production from the very beginning. The approach continues previous integrated design research at the Institute for Computational Design (ICD) and the Institute of Building Structures and Structural Design (ITKE) on fibre-based structures with a mould-less winding technique. This is discussed through the implementation of a full-scale architectural prototype with a pneumatic formwork that reduces the necessary formwork for fibre composite to the minimum and serves after fabrication as envelope. The paper focuses on the project-oriented integrated design strategies, including the challenges and the implementation of new process-specific computational tools.
We present a novel, deeply embedded robotics middleware and programming environment. It uses a multithreaded, publish-subscribe design pattern and provides a Unix-like software interface for micro controller applications. We improve over the state of the art in deeply embedded open source systems by providing a modular and standards-oriented platform. Our system architecture is centered around a publish-subscribe object request broker on top of a POSIX application programming interface. This allows to reuse common Unix knowledge and experience, including a bash-like shell. We demonstrate with a vertical takeoff and landing (VTOL) use case that the system modularity is well suited for novel and experimental vehicle platforms. We also show how the system architecture allows a direct interface to ROS and to run individual processes either as native ROS nodes on Linux or nodes on the micro controller, maximizing interoperability. Our microcontroller-based execution environment has substantially lower latency and better hardware connectivity than a typical Robotics Linux system and is therefore well suited for fast, high rate control tasks.
Im Bereich des automatisierten Straßenbaus gab es in den letzten Jahren sehr große Fortschritte, sodass auf großen Autobahnbaustellen schon vielfach automatisierte Baumaschinen eingesetzt werden. Bei diesen Maschinen findet hauptsächlich nur eine automatische Höhenregelung der Werkzeuge statt. Eine vollständige automatische Steuerung von Lage und Höhe des Fahrzeuges und Werkzeuges ist nur bei langsam fahrenden Baumaschinen, wie z.B. bei Asphaltfertigern und Randsteinfertigern, umgesetzt. Der Aufbau solcher Systeme wird zumeist individuell für jede Maschine durchgeführt, wobei das Zusammenspiel von Sensoren, Filter- und Regelalgorithmen von zentraler Bedeutung ist und für jeden Baumaschinentyp erneut vorgenommen werden muss. Simulationen für die Implementierung der Software auf dem Maschinenrechner und den zu verwendenden Sensoren werden meist nur softwareseitig durchgeführt. Ein Zwischenschritt, bei dem sowohl einzelne Sensorkomponenten oder auch die zu implementierenden Filter im Labor ohne äußere Einflüsse, wie z.B. Bodenbeschaffenheit oder andere Umgebungseinflüsse, getestet werden können fehlt. Das in dieser Arbeit entwickelte dreistufige Simulationskonzept schließt mit der Entwicklung eines zusätzlichen Hardware-In-The-Loop Simulators diese Lücke. Für den Hardware-In-The-Loop-Simulator werden ferngesteuerte Fahrzeugmodelle im Maßstab 1:14 eingesetzt, welche die kinematischen Fahreigenschaften realitätsnah abbilden. Durch das Zusammenschalten von Sensoren, Software und den Modellen können entsprechende Simulationen durchgeführt werden, um die neu einzubindenden Sensoren und Algorithmen im Labor zu optimieren. Zur Umsetzung des Simulators findet, neben der geometrischen Beschreibung von weitverbreiteten Fahrzeugmodellen im Straßenbau, auch eine exemplarische Beschreibung von drei charakteristischen Werkzeugen statt. Bei diesen handelt es sich um die Werkzeuge einer Planierraupe, eines Motorgraders und eines Asphaltfertigers. Anhand von zwei dieser geometrischen Werkzeugmodelle wird eine Sensitivitätsanalyse durchgeführt, welche den Einfluss verschiedener Sensoren auf die erreichbare Positionsgenauigkeit der Werkzeuge untersucht. Zum besseren Verständnis der systemtheoretischen und regelungstechnischen Begriffe für die Beschreibung des Simulators wird eine kurze Einführung in diese Thematik gegeben. Dabei wird gezielt auf die im Simulator benutzten Regler zur Querregelung der Fahrzeugmodelle auf einer vorgegebenen Trajektorie (Soll-Trajektorie) eingegangen. Die verwendeten Regler bestehen aus verschiedenen Kombinationen der einzelnen Parameter des PID-Reglers und beruhen somit auf einem nicht modellbasierten Ansatz. Für die Beschreibung der Regelgüte, am Beispiel der Querregelung, wird zur Evaluierung der Sensoren sowie der Regel- und Filteralgorithmen im Regelkreis der quadratische Mittelwert (Root Mean Square = RMS) eingeführt. Dieser berechnet sich aus den Querabweichungen der Fahrzeuge zu einer Soll-Trajektorie während der Querregelung. Vor dem Einsatz der Regler im Simulatorsystem werden diese anhand von Softwaresimulationen evaluiert. Zur Positionsbestimmung der zu regelnden Fahrzeuge werden drei Robot-Tachymeter im Simulator eingesetzt. Diese werden hinsichtlich ihrer Messgenauigkeit für die Bestimmung der Regelgüte untersucht. Hierfür wird als Positionsreferenz zusätzlich ein Laser-Tracker verwendet, mit dessen Hilfe eine Trennung von Regelgüte und Messgenauigkeit erreicht werden kann. Bei den im Simulator integrierten Fahrzeugen handelt es sich um ein Raupen- und um ein LKW-Modell, mit deren Hilfe ein Großteil der im Straßenbau vorkommenden Fahrzeugmodelle untersucht werden kann. Für diese Modelle wird ein automatisiertes Kalibrierverfahren zur Bestimmung der Lenkparameter zur rechnergestützten Steuerung vorgestellt. Bei den praktischen Experimenten, die mit Hilfe des Simulators durchgeführt werden, findet neben der Überprüfung der implementierten Fahrzeugmodelle auch ein Vergleich von zwei Kalman-Filter-Varianten zur Verbesserung der Regelgüte statt. Innerhalb der Testreihen konnte in einem Geschwindigkeitsbereich von 10 - 30 cm/s eine Regelgüte von 2 - 4 mm erreicht werden.
The art installation Flight Assembled Architecture [1] is one of the first structures built by flying vehicles. Culminating in a 6-m-tall tower composed of 1500 foam modules (see Figures 1 and 2), the installation was assembled by four quadrocopters in 18 hours during a four-day-long live exhibition at the Fonds R?gional d'Art Contemporain (Regional Contemporary Art Fund) du Centre in Orl?ans, France. This article documents the design and development of specific elements of the autonomous system behind this one-of-a-kind installation and describes the process and challenges of bringing such a complex system out of the laboratory and into the public realm, where live demonstration and human-in-the-loop interaction demand high levels of robustness, dependability, and safety. The installation is a 1:100 scale model of what was originally conceived of as a 600 m-high vertical village (see "The Vertical Village" for details) and is an exploration of aerial construction in architecture. Architects have been exploring the use of digital technologies for the design and assembly of structures for some time now, and many facilities for investigating nonstandard architectural design and fabrication using industrial robots have sprung up in the past decade [2]-[4]. However, robot arms and computer numerical control (CNC) machines are limited by predefined working areas that constrain the size of the workpiece they can act upon and are thus also limited in their scale of action to a small portion or component of the overall structure, or to model-sized fabrication [5]. In contrast, flying machines are not constrained by such tight boundaries. The space that flying machines can act upon is substantially larger than the size of the machines themselves, making it feasible for the machines to work on the structure as a whole at a 1:1 scale, thus offering architects a new framework for realizing their designs.
This article provides a tutorial introduction to modeling, estimation, and control for multirotor aerial vehicles that includes the common four-rotor or quadrotor case.