DESIGN AND TESTING OF A LOW-COSTFLIGHT CONTROL AND DATA ACQUISITION SYSTEM

FOR UNSTABLE SUBSCALE AIRCRAFT

Alejandro Sobron , David Lundström , Ingo Staack , Petter KrusLinköping University, Linköping, Sweden

Keywords: subscale, flight testing, flight control, data acquisition, relaxed stability

Abstract

Current research on subscale flight testingmethodologies at Linköping University is per-formed by using various platforms, some of themwith advanced configurations. These have beenpreviously flown in open-loop under direct com-mands from the pilot. However, the interest inflying some of these platforms with relaxed sta-bility and the investigation of multi-surface con-trol allocation techniques motivated the imple-mentation of a simple low-cost flight control sys-tem based on commercial-off-the-shelf compo-nents. The work described in this paper evaluatesthe simplest available solutions that provide con-trol augmentation for small, longitudinally unsta-ble, free-flying models. This work also tries todefine a reliable, fail-safe system architecture thatcan be implemented in more advanced platforms.Moreover, data acquisition and analysis are eval-uated with the aim of applying system identifica-tion techniques.

1 Introduction

Testing physical subscale models has always con-stituted a valuable tool in aircraft development.Despite recent advances of modern computa-tional techniques, wind tunnel testing still con-stitutes the backbone of aircraft design and aero-dynamic research. However, free-flying subscalemodels have proved an excellent complement,and in some cases a lower-cost alternative tothese techniques. This is especially interestingfor high-risk conditions and for evaluation of un-conventional designs for which no previous expe-

rience and simulation models exist; see for exam-ple NASA’s experiences summarised by Cham-bers in [1].

The miniaturisation of complex mechatronicsystems and the development of rapid proto-typing technologies have lowered significantlythe cost and the manufacturing time of smalland mid-size remotely-piloted aircraft systems(RPAS). Further, miniaturised electronics allowto obtain precise quantitative measurements ofthe aircraft properties, in addition to the usualqualitative assessment. These new possibilitiesmake subscale flight testing an attractive evalua-tion tool to be integrated in the design loop dur-ing the initial phases of full-scale aircraft devel-opment.

It has also become affordable for aca-demic institutions where it additionally servesas an extraordinary practical training for stu-dents [2]. Current research on subscale flighttesting methodologies at Linköping University(LiU) is performed by using various platforms,some of which feature advanced configurationssuch as the Generic Future Fighter (GFF) sub-scale demonstrator, described in [3] and shownin Fig.1. Although equipped with data acquisi-tion systems [4], these platforms have been previ-ously flown in open-loop in stable configurations.However, the interest in flying some of them withthe relaxed stability configuration they were orig-inally designed for, as well as the possible inves-tigation of multi-surface control allocation tech-niques, motivated the implementation of a sim-ple, low-cost flight control system (FCS) [5].

1

SOBRON A, LUNDSTRÖM D, STAACK I, KRUS P

The work presented here tries to evaluatethe simplest available commercial off-the-shelf(COTS) solutions that can provide control aug-mentation even for the high-frequency, short-period dynamics of small free-flying models of-ten used in student projects, Fig.2. Autonomousnavigation capabilities are disregarded. Thiswork also tries to define a reliable, fail-safe sys-tem architecture to be implemented on the moreadvanced platforms. Furthermore, integrateddata acquisition possibilities are evaluated withthe aim of analysing the aircraft characteristicsand applying system identification techniques.

Fig. 1 . GFF subscale research platform, described in [3].

Fig. 2 . Some of the small test-bed models used for experi-mental verification of equipment and procedures.

2 Frame of Reference

The aim is to integrate COTS solutions wherepossible rather than developing a new systemfrom scratch. Disregarding autonomous flightfunctions of the so called “autopilots”, there arecurrently different solutions that provide aug-mented flight control to small RPAS. Ground-based FCS solutions process downloaded sensor

data and upload direct control surface commands,such as the system used by NASA in [6]. How-ever, this system relies upon a solid data trans-mission link with low latency which would bedifficult to achieve with a limited budget. There-fore, a FCS onboard the aircraft, following theconcept sketched in Fig.3 is preferred despiteeventual limitations in processing power.

Signal processingand digital modulation

Transmission

ReceptionSignal modulation

FCS

Sensors

Signal mixing

Servo-mechanism

Control Laws

Control surfaceControl stick

Visual feedback

Stability and ControlAugmentation

Systemfeedback

Fig. 3 . Concept diagram of a RPAS with a FCS onboard.

A basic stability augmentation system (SAS)can be achieved by using relatively simple andinexpensive equipment only: angular rate closed-loop controllers based on single-axis gyroscopesensors. These are extensively used in the radio-controlled (R/C) market to stabilize the direc-tional axis (tail rotor pitch) of R/C helicopters.Applied on aeroplanes, the authors have previousexperience using the MEMS gyroscope FutabaGYA352 [7] to stabilize roll and pitch in oneof the jet-powered research platforms. However,only vague examples of its use in highly unstableconfigurations were found. Although this solu-tion would require an external data logger, it wasdecided to explore further this concept for its pos-sible use as a back-up system.

Regarding more advanced control, Cai et al.present in [8] a thorough review of FCS used indifferent kinds of UAVs. According to these au-thors, PC/104 embedded computers are widelyused in mid-size platforms such as in [9]. How-ever both cost and form factor are still exces-sive for this application. Dantsker et al. [10] of-fer a comprehensive review of several FCS withsmall form factor that also log and process sensordata, both closed-source and open-source COTSproducts as well as custom-made avionic sys-tems such as the one developed by Beard et al.[11]. Open-source solutions are favoured notonly for cost reasons, but also for flexibility of

2

DESIGN AND TESTING OF A LOW-COST FLIGHT CONTROL AND DATA ACQUISITION SYSTEMFOR UNSTABLE SUBSCALE AIRCRAFT

development. For example, the platform Ardupi-lot Mega has been used as a base for similarprojects such as in Hartley et al. [12] and in Ar-ifianto and Farhood [13]. Nevertheless, amongthe current options two platforms based on 32bit microprocessors with sensor data process-ing and logging capabilities stood out becauseof their promising performance and connectiv-ity: Paparazzi firmware combined with Apogeev1.00 hardware [14], and PX4/APM firmwarecombined with Pixhawk hardware [15].

The Pixhawk, a project started by theComputer Vision and Geometry Lab in ETHZürich, supported by the Linux FoundationDronecode community and the private company3D Robotics, was finally selected for further in-vestigations. Besides an embedded NuttX real-time operating system (RTOS) and a specificallydeveloped PX4 middleware layer, the user canchoose the flight control firmware to be run ontop: PX4 autopilot from the same authors, or thewell-known APM multiplatform autopilot, bothopen-source too. Due to the larger developerscommunity and better documentation, the fixed-wing version of the APM firmware, known asAPM-Plane [16] was chosen here as a startingpoint for further development.

Despite of the already notable capabilities ofthese high-end hobbyist autopilot systems, in theopinion of the authors it is expected that futuredevelopments will be increasingly based on stan-dardised real-time Linux-based platforms, suchas Navio2 and Raspberry Pi [17].

3 System Design

The aim is to assemble a compact, modular sys-tem that could be exchangeable between differenttest-bed platforms. Critical requirements wereminimum physical size, weight, and power con-sumption for the main flight controller as well asthe sensors and connected peripherals. In addi-tion, a fail-safe design needed to be considered:even if the FCS can switch to direct manual con-trol in case of system failure, the pilot would notbe able to control an unstable aircraft. Therefore,some alternative must be provided, and only twopossibilities exist: modifying the static margin

in flight, or engaging a secondary SAS. The firstoption implies either moving the centre of grav-ity forward using a weight shifting mechanism ormoving the neutral point backwards, which in thecanard-delta configuration can be done by align-ing the canards with the airflow. A weight shift-ing mechanism with sufficient effect on the bal-ance was considered difficult to integrate, sim-ilarly to some mechanism for freeing and re-attaching the canards in place. Using direct in-formation from an angle of attack (AOA) vanesensor to command the canards would still re-quire functional sensor data processing. Instead,an alternative force-sensing control law for theservo actuators was conceived and investigated:an R/C servo was modified and analysed in a testrig, but results were not satisfactory due to com-plicated sensitivity adjustment and excessive hys-teresis effects. For these reasons, the option ofplacing an auxiliary R/C MEMS gyroscope con-troller providing a basic SAS was further inves-tigated using small electric-powered test-bed air-craft both in delta-canard and conventional con-figurations as in the example shown in Fig.4.

MEMS gyro sensor

High-speedservo actuator

Variable CG position

GY

Fig. 4 . A MEMS gyroscope sensor was installed in a Mul-tiplex FunCub model in order to test a basic longitudinalSAS approach with negative static margin.

Some remarkable characteristics of the se-lected hardware, the Pixhawk controller [15], areits good connectivity and the integrated capa-bility of logging sensor data at acceptable sam-pling rates, as shown in Fig.5. The APM-Planefirmware code was modified in order to increasethe logging rate of some variables of interest,such as the pilot commands.

However, the logging performance is stilllimited and any further increase in sampling rateswould be better managed with an external datalogger device.

3

SOBRON A, LUNDSTRÖM D, STAACK I, KRUS P

Main ProcessorSTM32F427

32 bit, 168 MHz

Safety ProcessorSTM32F103

32 bit, 72 MHz

Accelerometers 1000+ Hz

Gyroscopes 1000+ Hz

Magnetometer 130 Hz

Barometer 130 Hz

Alpha Beta vanes 50+ Hz

Airspeed sensor 10+ Hz

GPS module 5 Hz

Pilot commands 50 Hz

50 Hz

50 Hz

10 Hz

10 Hz

50 Hz

10 Hz

5 Hz

50 Hz

Data LoggerMicro SD Card

controls

controls

controls

telemetry

PWM

CA

N bus

I2C bus

UA

RT

INPUTS PROCESSING LOGGER

OU

TPUTS

...

Fig. 5 . Main I/O of the FCS and respective sam-pling/logging rates after modification of the APM-Planefirmware. Adapted from an original diagram by the PX4developers community [15].

3.1 Instrumentation

The sensors already in-built in the Pixhawk boardare inexpensive and not certified for professionalflight testing, but have proved robust in harshenvironments with vibration and large temper-ature variations such as those usually found inhobbyist-type operations. The main board com-prises two different inertial measurement units(IMU) with an integrated magnetometer ([18],[19] and [20]), and a barometric pressure sen-sor [21]. Additional external sensors can be inte-grated to complete the desired FCS architecture,as the example shown in Fig.6. The most relevantperipheral components are discussed below.

3.1.1 Airspeed sensorPrevious experiences with analogue pressuretransducers from other hobbyist-type systemsshowed that these suffer severe variance withtemperature. Here, inexpensive COTS pressuretransducers with digital output and temperaturecompensation 3D Robotics are used. Since anappropriate wind tunnel was not available, cali-bration was done in flight as explained in [5].

3.1.2 Airdata boomVarious airdata booms were built in-house com-bining a pitot-static probe with flow-directiontransducers. Fig.7 shows the integration ofthe pitot-static probe with a 3D-printed AOA

G

F R

E

T

P

A

A

F

G

B

S

R

T

E

M

Alpha vane

Beta vane

Voltage + current sensor

Flight controller (+ 2xIMU + baro.)

GPS module + magnetometer

Motor

Radio link receiver

Servo

Telemetry modem

B

O

C

L

Y

M

S S

C Camera

L Dedicated LiPo battery

O OSD + video recorder

P Pitot tube + Termometer

Y Auxiliar gyroscope sensor

M

S

E

R

G

L

T

P

A

SS

F

Fig. 6 . Layout of the proposed FCS and its main sensors inthe GFF platform and one of the smaller test-bed models.

vane in the lightest airdata boom used inthe small test-bed aircraft. A larger boomis shown in Fig.8. Angles are measuredusing magnetic-induction rotary encoders ex-tracted from inexpensive hobbyist-type R/C ser-vos HK28013DMG. The angular position is com-puted by measuring the linear analogue output ofthe encoders through two of the available ADCports on the Pixhawk board.

3.1.3 GPS and external magnetometerElectromagnetic noise caused by onboard equip-ment can disturb sensible sensors and receiverantennas, as measured by Dantsker et al. [10].It is desirable to place a remote GPS antennaand a secondary magnetometer as far as possi-ble from the main electric devices. In this case,the recommended low-cost COTS module manu-factured by 3D Robotics was used. It integratesa U-blox GPS module [22] and a digital magne-tometer [23] in a small and light package.

4

DESIGN AND TESTING OF A LOW-COST FLIGHT CONTROL AND DATA ACQUISITION SYSTEMFOR UNSTABLE SUBSCALE AIRCRAFT

Fig. 7 . CAD model of the small airdata probe (8 mm ofdiameter) combining a pitot-static tube and a flow-directiontransducer. The sensor housing, the vane arm, and the vanewere 3D-printed in ABS plastic.

Fig. 8 . Larger airdata nose-boom with alpha and betatransducers installed in the GFF platform.

3.1.4 Voltage and current sensorPower sensors typically measure the main and ra-dio battery levels in hobbyist rigs. Here this set-up is only used in the smaller electric-poweredtest-bed aircraft. On the jet powered platforms, itmonitors the fuel pump activity in order to mea-sure the fuel consumption.

3.1.5 Video camera and OSDLight micro-cameras, with or without on-screen-display (OSD), capture phenomena of interestsuch as the performance of the flow-angle trans-ducers (Fig.11) or the attitude of the aircraft inreference to the horizon.

4 Modelling and Simulation

A simulation model is the base for evaluatingthe flight dynamics of the flying platforms aswell as the effect of control laws, control looptuning, and later on for the application of sys-tem identification techniques. Following a state-space approach, two levels of complexity are de-fined: a simplified, approximated linear model ofthe small test-bed aircraft and a more exhaustivenon-linear model of the research platforms. Thelatter is not only based on theoretical and numer-ical analyses, but also on experimental data suchas wind tunnel tests. It is still under developmentat the time of writing this paper.

4.0.1 Mass and inertia propertiesGiven the size of the aircraft it is easy to mea-sure accurately the total mass and that of thecomponents onboard. The inertia properties ofthe research platforms was measured experimen-tally using the pendulum motion technique [24].However, due to the difficulties of doing this onthe very light airframes, the inertia properties ofthese aircraft were estimated using a MATLABscript based on CAD airframe data and discretepoint loads, Fig.9.

Fig. 9 . One of the lightweight test-bed aircraft modelled inthe MATLAB script developed to estimate numerically theinertia properties.

4.0.2 Aerodynamic modelPreliminary aerodynamic coefficients were esti-mated using the vortex lattice method (VLM)Tornado [25]. A traditional drag bookkeepingmethod supported by CAD wetted area compu-tation was used to estimate the parasite drag. A

5

SOBRON A, LUNDSTRÖM D, STAACK I, KRUS P

first estimation of the neutral point location wasobtained from the VLM computation, althoughthis was later verified and refined through flighttesting. Propeller effects were disregarded.

4.0.3 Servo actuatorsThe servo actuator model is particularly impor-tant. One of the main problems with the inte-grated data acquisition system is that it lacks suf-ficient analogue inputs to measure the real con-trol surfaces deflection. Therefore the actual de-flection has to be estimated from the commandedpulse-width modulated signal using an accuratemodel of each servo. The dynamics are assumedto be a second order, critically damped systemwith rate limit [12]. By experiment, the max-imum rate and bandwidth for every servo weremeasured by comparison of a sinusoidal inputreference signal with the servo arm position ina simple test-rig.

4.0.4 FCSGaussian noise is added to the sensor measure-ments, but no further errors or bias. Exceptfor the nose-boom, no dynamic effects are in-troduced since the short distances to the cen-tre of gravity can be neglected. Alpha and betameasurements are corrected with the appropri-ate response equations taking into account thefast angular motions of the aircraft. The controllaws of the gyroscope controller and those avail-able by default in the APM-Plane [16] firmwarewere interpreted from the code and modelled ac-cordingly. These are simple PID control loopswith an airspeed scaling factor, available for bothpitch/roll attitude and angular rate, as in the ex-ample Fig.10.

+-

++

AircraftDynamicsServoAmp.

IMUrate

ec δEScalerX

+

1s

Int.

ScalerX+

-

k P

k I

k D

VVref

Airspeed Scaler

θ = qref.

ref θ = q.

Fig. 10 . Structure of the pitch rate control law of the APM-Plane firmware, version 3.2.1.

4.0.5 Engine ModelThe propulsion model is especially relevant forthe longitudinal motion dynamics. In small elec-tric motors, dynamics such as spool up time arenegligible compared to the rigid body dynamics[12]. However, this is taken into account in the jetengines by means of a delay roughly based on ob-servations. If existing, moments due to misalign-ment are taken into account, but the effects of thepropeller are disregarded. Static thrust was mea-sured directly, but the effects of airspeed are mod-elled according to traditional equations found inliterature.

5 Initial Flight Testing

An initial flight test campaign was focused oncalibrating the different sensors and verifyingtheir consistency and robustness. Figures 11 and12 illustrate part of this process with the airdataboom installed in the small test-bed aircraft.

Fig. 11 . Evaluation of the small AOA transducer perfor-mance and calibration by comparing logged flight data withvisual references captured by a micro-camera.

Previous unsatisfactory experiences with ahigh-end attitude and heading reference systems(AHRS) onboard small subscale aircraft showedthat a capable and robust algorithm is essen-tial for accurate attitude estimation in this en-vironment of high g-forces and rapid motions.The APM-Plane firmware, from version 3.5 on-wards, runs in parallel two instances of an ex-tended Kalman filter (EKF) taking advantage oftwo IMU data. The algorithm evaluates their con-dition in real time and selects the most consistent.To ensure robustness, a simpler direction-cosine-matrix algorithm runs in the background and it

6

DESIGN AND TESTING OF A LOW-COST FLIGHT CONTROL AND DATA ACQUISITION SYSTEMFOR UNSTABLE SUBSCALE AIRCRAFT

0 5 10 15 20 25−60

−50

−40

−30

−20

−10

0

10

20

30

40

Frequency (Hz)

Pow

er/F

requ

ency

(dB/

Hz)

Raw data Low−pass o�−line �lter, 2 Hz corner frequency

Fig. 12 . Power spectral density using fast Fourier trans-form, applied to the smallest AOA transducer to design anadequate filter.

is called if the two primary EKF fail. The con-sistency of the attitude and estimated trajectorywere tested in continuous high-g turns and otheraggressive manoeuvres using various platforms.As the example in Fig.13, test results suggest thatthe estimation is stable and fairly accurate.

Fig. 13 . One of the jet-powered platforms performing sus-tained high-g turns to test the robustness of the attitude es-timation. Flight data are superimposed on real video usingthe open-source software Mission Planner [26].

5.1 Tests with Relaxed Stability

The FCS was tested under relaxed stability con-figurations using the small test-bed aircraft. Al-though simulations gave an initial guess of thecontroller gains, the system was experimentally

tuned while the static margin was gradually re-duced. As an example of the effect of thestatic margin in the short period dynamics, Fig.14shows the response to pitch angle step inputsof the Multiplex FunCub (4) model at differentstatic margins with the same PID gains setting.Similar effects were observed with the pitch ratecontrol law. Stability was recovered in most casesby increasing the derivative (D) gain. Negativestatic margins in excess of minus 20 percent wereachieved successfully. Beyond this value, the re-sponse time of the pitch servos and the avail-able elevator authority became the limiting fac-tors. In some platforms, effects of lateral insta-bility started to appear before reaching this point.

Demanded pitch Actual pitch

(a)

(b)

(c)

−50

0

50A

ngle

[deg

]

−50

0

50

Ang

le [d

eg]

−50

0

50

Time [s]

Ang

le [d

eg]

0 5 10 15 20 25 30 35 40

Fig. 14 . Response to pitch angle step inputs of the Cubtest-bed with pitch angle control engaged and the same PIDgains setting, for: (a)8 percent of positive static stabilitymargin, (b)neutral static stability, (c)10 percent of negativestatic stability.

The single-axis gyroscope controller ap-proach was tested with an inexpensive hobbyist-type gyroscope Assan GA250, using both rateand angle-hold control laws. Results were sur-prisingly similar, although the lack of variablegains according to airspeed makes the tuning pro-cess more critical. In addition, since the gyro-scope reacts equally to any angular velocity aboutthe pitch axis, banked turns become excessivelydamped. However, similar values of negativestatic margin were achieved with this solutionand the same limiting factors were detected.

7

SOBRON A, LUNDSTRÖM D, STAACK I, KRUS P

5.2 Assessment of Manual Control

The small test-bed aircraft were also used to as-sess qualitatively the limits of manual controlwith the FCS disengaged. The observations sug-gest that the degradation of the flying qualitiesis highly airframe-dependent, and other factorssuch as wind speed and turbulence also affectthe outcome. Most platforms could be flownmanually up to the neutral static stability mar-gin, where they present significantly degradedflying qualities that could be rated as level 8-9 inthe Cooper-Harper scale [27]. It was possible tocontinue further into negative static margins withcertain platforms. However, assuming an averageskilled pilot, results suggest that it is not advis-able to go beyond minus five percent in any ofthe cases, see Fig.15.

0 2 4 6 8 10 12 14−50

−40

−30

−20

−10

0

10

20

30

40

50

Time [s]

Ang

le [d

eg]

Pitch rate, q Pitch angle, θ Angle of attack, α

−250

−200

−150

−100

−50

50

100

150

200

250

0

FCS disengaged

Ang

ular

rate

[deg

s−1

]

Fig. 15 . Uncontrollable attitude oscillations occur whentrying to control manually the Cub model with a negativestatic margin of -10 percent, with the FCS disengaged.

6 Discussion

The presented low-cost FCS architecture is basedprimarily on a COTS system, which apart fromthe flight controller, integrates a sensor data log-ger. Different control laws have been flight-testedunder relaxed stability conditions in diverse smalltest-bed aircraft with successful results. Observa-tions suggest that the main limiting factors for theeffective control of highly unstable platforms arethe delay of the pitch servo actuators and the au-thority of the pitch control surfaces at low speeds.The system has also been evaluated in open-loopsetup on the larger jet-powered platforms.

The observed difficulties to maintain manualcontrol under relaxed stability settings suggestthat it is critical to incorporate a secondary SAS

in case of a FCS failure. Flight tests have provedthat a simple single-axis gyroscope controller isa valid and inexpensive solution.

The proposed setup also works as a data ac-quisition system. Appropriate filters have beendesigned and the original firmware has beenmodified to increase the sample rate of someparameters. Accurate trajectory reconstruction(Fig.16) is easily obtained, and attitude estima-tion seems robust and consistent according to thetest results.

A MATLAB script has been developed in or-der to display, analyse and pre-process the flight-test data for application of system identificationtools. The complete simulation model of the GFFresearch platform is currently under developmentand due to time limitations, system identifica-tion results could not be included in this publi-cation. However, the data obtained already of-fer a preliminary quantitative assessment of theflight characteristics of the GFF platform, as inthe example shown in Fig.17 for a high angleof attack manoeuvre. However, to increase fur-ther data acquisition capabilities it is proposed tocomplement the FCS with a standalone computeror a dedicated data logger.

Fig. 16 . Trajectory reconstruction of a flight test exportedinto Google Earth using the open-source software MissionPlanner [26].

7 Conclusion

The system presented here is a valid low-costsolution for flight control of subscale platforms

8

DESIGN AND TESTING OF A LOW-COST FLIGHT CONTROL AND DATA ACQUISITION SYSTEMFOR UNSTABLE SUBSCALE AIRCRAFT

0 5 10 15 20 25 30Time [s]

-20

-10

0

10

20

30

40

50

60

Ang

le [d

eg]

Velo

city

[m s

-1]

True airspeed Vertical speed Pitch, θ AOA, α

Fig. 17 . High angle of attack manoeuvre performed by theGFF research platform.

with relaxed stability. Its modular architec-ture based almost entirely on open-source COTScomponents requires a minimum developmenttime and allows any desired modifications. Boththe primary FCS and the backup system havebeen tested in different small test-bed aircraft andtheir satisfactory performance seems only limitedby the control surface actuation.

As a data acquisition system, it provides dataof sufficient quality to allow identification offlight dynamical characteristics. Specific soft-ware tools have been developed to visualise andpre-process the flight-data for further analysis,but at the time of writing this paper the completesimulation models and the system identificationtools are still under development.

References

[1] Chambers J R. Modeling Flight: The Role ofDynamically Scaled Free-Flight Models in Sup-port of NASA’s Aerospace Programs. 1st ed.NASA, Washington, DC, USA, 2009.

[2] Jouannet C, Berry P and Krus P. Aircraft DesignEducation at Linköpings University. Proceed-ings of the Institution of Mechanical Engineers,Part G: Journal of Aerospace Engineering, vol.221, pp. 217–224. 2007.

[3] Amadori K, Jouannet C and Berry P. Develop-ment of a subscale flight testing platform for ageneric future fighter. 27th Congress of Inter-national Council of the Aeronautical Sciences.Nice, France, 2010.

[4] Staack I and Lundström D. Subscale FlightTesting at Linköping University. 27th Congressof International Council of the AeronauticalSciences. Nice, France, 2010.

[5] Sobron A. Design and Testing of a FlightControl System for Unstable Subscale Aircraft.Master’s thesis, Linköping University, 2015.

[6] Jordan T L, Foster J V, Bailey R M and Bel-castro C M. AirSTAR: A UAV Platform forFlight Dynamics and Control System Testing.25th AIAA Aerodynamic Measurement Technol-ogy and Ground Testing Conference. AmericanInstitute of Aeronautics and Astronautics, SanFracisco, CA, USA, 2006.

[7] Futaba. GYA352 MEMS Gyroscope, 2016.URL https://www.rc.futaba.co.jp/

english/gyro/airplane/gya352

[8] Cai G, Dias J and Seneviratne L. A Survey ofSmall-Scale Unmanned Aerial Vehicles: RecentAdvances and Future Development Trends. Un-manned Systems, vol. 02, no. 02, pp. 175–199,2014.

[9] Jang J S and Tomlin C J. Design and Implemen-tation of a Low Cost, Hierarchical and ModularAvionics Architecture for the DragonFly UAVs.AIAA Guidance, Navigation, and Control Con-ference and Exhibit, August, pp. 4465–4477.AIAA, Monterey, CA, USA, 2002.

[10] Dantsker O D, Mancuso R, Selig M S and Cac-camo M. High-Frequency Sensor Data Ac-quisition System (SDAC) for Flight Controland Aerodynamic Data Collection. 32nd AIAAApplied Aerodynamics Conference. AIAA, At-lanta, GA, USA, 2014.

[11] Beard R W, Kingston D, Quigley M, SnyderD, Christiansen R, Johnson W, McLain T andGoodrich M. Autonomous Vehicle Technolo-gies for Small Fixed-Wing UAVs. Journal ofAerospace Computing, Information, and Com-munication, vol. 2, no. 1, pp. 92–108, 2005.

[12] Hartley R F, Hugon F D X, Anderson R andMoncayo H. Development and Flight Testingof a Model Based Autopilot Library for a LowCost Unmanned Aerial System. AIAA Guid-ance, Navigation, and Control (GNC) Con-ference, pp. 1–25. AIAA, Boston, MA, USA,2013.

[13] Arifianto O and Farhood M. Development andModeling of a Low-Cost Unmanned Aerial Ve-

9

SOBRON A, LUNDSTRÖM D, STAACK I, KRUS P

hicle Research Platform. Journal of Intelligent& Robotic Systems, vol. 80, no. 1, pp. 139–164,2015.

[14] École nationale de l’aviation civile and Pa-parazzi develpers community. Paparazziproject, open-source autopilot system, 2015.URL http://wiki.paparazziuav.org/

[15] Computer Vision and Geometry Lab at ETHZürich 3D Robotics Inc and PX4 developerscommunity. PX4/Pixhawk open-hardware au-topilot project, 2015.URL http://pixhawk.org/

[16] Dronecode Project Inc a Linux Foundation Col-laborative Project. APM:Plane open-source au-topilot firmware, 2016.URL http://plane.ardupilot.com/

[17] Emlid. Navio2 Linux autopilot, 2016.URL https://emlid.com/

introducing-navio2/

[18] STMicroelectronics. LSM303D Three-axisMEMS accelerometer and magnetometer, 2013.

[19] STMicroelectronics. L3GD20 Three-axisMEMS digital gyroscope, 2013.

[20] InvenSense. MPU-6000 Six-axis MEMS motiontracking device, 2013.

[21] Measurement Specialities. MS5611 Baromet-ric pressure sensor. Measurement Specialties,2012.

[22] U-blox. NEO-7 GPS/GNSS module, 2014.[23] Honeywell. HMC5883L Three-axis digital

magnetometer, 2013.[24] Jordan T, Langford W and Hill J. Airborne

Subscale Transport Aircraft Research Testbed- Aircraft Model Development. AIAA Guid-ance, Navigation, and Control Conference andExhibit. American Institute of Aeronautics andAstronautics, San Fracisco, CA, USA, 2005.

[25] Melin T. Tornado, the Vortex Lattice Method,Version 135, 2010.URL http://www.redhammer.se/tornado/

[26] Oborne M and Dronecode Project Inc. Mis-sion Planner open-source ground station soft-ware, Version 1.3.37, 2016.URL http://ardupilot.org/planner/

[27] Nelson R C. Flight stability and automatic con-trol. 2nd ed. WCB/McGraw-Hill, Singapore,1998.

Contact Author Email Addressalejandro.sobron@liu.se

Copyright StatementThe authors confirm that they, and/or their companyor organization, hold copyright on all of the origi-nal material included in this paper. The authors alsoconfirm that they have obtained permission, from thecopyright holder of any third party material includedin this paper, to publish it as part of their paper. Theauthors confirm that they give permission, or have ob-tained permission from the copyright holder of thispaper, for the publication and distribution of this pa-per as part of the ICAS proceedings or as individualoff-prints from the proceedings.

10