Accuracy of a Low-cost Autonomous Hexacopter Platforms Navigation Module for a Photogrammetric and Environmental Measurements

Transcription

1 Environmental Engineering 10t International Conference eissn / eisbn Vilnius Gediminas Tecnical University Lituania, April 2017 Article ID: enviro ttp://enviro.vgtu.lt DOI: ttps://doi.org/10.86/enviro Accuracy of a Low-cost Autonomous Hexacopter Platforms Navigation Module for a Potogrammetric and Environmental Measurements Paweł Burdziakowski 1, Katarzyna Bobkowska 2 Faculty of Civil and Environmental Engineering, Gdansk University of Tecnology, Gdansk, Poland 1 pawel.burdziakowski@wilis.pg.gda.pl; 2 katarzyna.bobkowska@wilis.pg.gda.pl Abstract. A potogrammetry and environmental measurements from an unmanned aerial veicle (UAV) are a low-cost alternative for a traditional aerial potogrammetry. A commercial off-te-self products (COTS) offers a variety of ceap components tat a suitable to be used on board a UAV. In tis paper a low-cost navigation module based on Ublox NEO-M8N GPS and Pixawk fligt controller ave been described, as a main extrinsic parameters source for aerial potogrammetry, mounted on autonomous exacopter platform. An accuracy of navigation data logged during te drone fligt by a fligt controllers navigation module as been compared wit geodetic reference measurement. A man controlled fligt modes as been tested during a drones loitering above designated point. Due to te situation tat, a variety of active electronic, radio transmitters, navigation, inertial and magnetic sensors, are to be placed on a relatively small exarotor platform wit ig current power system, overall interference between electronics ave to be considered. Paper presents results of navigation module fine tuning process and some tecnical issues solved during researc, tat caused an accuracy improvement and interference diminising. Detailed results are presented and concluded as for te accuracy of navigation based on ceap GPS module is satisfied for execution an autonomous potogrammetry and environmental tasks. Keywords: drone, uav, potogrammetry, navigation, open-source, pixawk. Conference topic: Tecnologies of geodesy and cadastre. Introduction Todays, a potogrammetry and environmental measurements from a unmanned aerial veicle (UAV) are a low-cost alternative for a traditional aerial potogrammetry. Low-cost satellite navigation tecnologies available on market are suitable to be used on board aerial veicles. A commercial UAVs can be are defined as a commercially available and manufactured by a one company complete product designed and optimized for aerial potograpy and videograpy. Te commercial UAVs are using one frequency GPS modules to calculate veicle geograpical position in order to support loiter flying mode. In loiter mode veicle automatically attempts to maintain te current location, eading and altitude. In tis mode te pilot may fly te veicle as if it were in manual fligt mode, but wen te control sticks are released, te veicle will slow to a stop and old position. Tis means tat, te GPS coordinates are used to maintain veicle D position in open environment. For tis fly mode a GPS satellite signal reception is crucial, and any satellite signal loss will unable to veicle to take-off safely and enter in loiter mode. On board non-commercial UAV, a GPS readings can feed different navigation systems, especially autonomous navigation patc planning and execution. Te UAV Te tested UAV platform (Burdziakowski 2017) is equipped wit two GPS modules. Te main module () is placed on carbon mast, 12 cm above top frame plate (Fig. 1a). Te was moved upwards to diminis electromagnetic interference (EMI) form ESCs (electronic speed controller), video and telemetry transmitters (placed between a top and bottom main frame plate) and main power ig current cables. Te second () is placed above a CCD sensor on te same level as te top frame plate. Te as been modified. A standard PCB 25 mm/2mm ceramic GPS antenna (ANT A) as been removed and new PCB ceramic antenna 25mm/mm (ANT90-250A) as been soldered. Tis modification allows to receive GLONASS satellite system signal (antenna is tuned for GPS and GLONASS frequencies) (Table 1). Bot GPS modules (UBlox M8N) ave been activated to receive GPS and GLONASS satellite navigation systems. FC (fligt controller) as been configured to receive satellite position from bot receivers, but for navigation system only source wit lower HDOP value is selected by FC algoritm. During stationary tests te modified receiver () received signal averagely form more satellites tan standard one. Due to tat reason te was mounted wit closer distance to electromagnetic source on board UAV Paweł Burdziakowski, Katarzyna Bobkowska. Publised by VGTU Press. Tis is an open-access article distributed under te terms of te Creative Commons Attribution (CC B-NC.0) License, wic permits unrestricted use, distribution, and reproduction in any medium, provided te original autor and source are credited.

2 Table 1. GPS antennas caracteristics (TME-Electronic) Items ANT A () ANT90-250A () Nominal frequency MHz 75.2± ~1608 (GPS: 75, GLONASS : 98~1608) Center frequency MHz 75± ±.0 VSWR at Fo max Return Loss at 85 MHz max db Return Loss at 1610 MHz max db Impedance(Ω) Polarization Model RHCP RHCP Frequency Temperature Coefficient 20ppm/deg. max 20ppm/deg. max A test fligt sows a differences between number of satellites (NSats) received and a HDOP between bot GPS receivers (Fig. 1b). Te was affected by EMI (electromagnetic interference) and received less number of satellites. A ig buildings in test area covers satellite signal reception on lower altitude. At te iger altitude bot receivers NSat (number of satellites) increases and HDOP (orizontal dilution of precision) is decreasing. 100 Heigt Baro GPS 0 Loiter fly mode Fig. 1. GPS modules and EMI sources localization on board UAV (a), number of satellites and HDOP in eigt domain (b) NSats HDOP 0.8 a) b) GPS modules are calculate veicle geograpical position in order to support a loiter flying mode. In te loiter mode veicle maintains te current location, eading and altitude desired position. All sensors inputs are calculated using EKF (Te Extended Kalman Filter). Te GPS position is a part of all inputs. Te veicle position (calculated using EKF) can be different tan pure GPS readings. Te EKF algoritm uses gyroscopes (GRO), accelerometer (AC- CEL), 2 compass (COMP), 2 GPSs, and 1 barometric (BARO) pressure measurements to estimate te position, velocity and angular orientation of te veicle. Wen te GPS measurement arrives, te filter calculates te difference (innovation) between te predicted position from inertial part and te position from te GPS. Te innovation, state covariance matrix, and te GPS measurement error are combined to calculate a correction to eac of te filter states (state correction). Te used ArduCopter firmware (v..) uses a 2 state EKF to estimate te following states (Ardupilot 2016): Attitude (Quaternions), Velocity, Position, Gyro bias offsets, Gyro scale factors, Z accel bias, Eart magnetic field, Body magnetic field, Wind Velocity. Te veicle orizontal position is delivered by te GPS and INS modules, and tis is considered as a main source of orizontal position (Fig. 2). Te veicle vertical position (altitude) is delivered by a barometer sensor and INS Z and GRO Z calculation. Te GPS eigt is not considered as an altitude reliable source, due to a reading fluctuations. Fig. presents altitude reading from a different source during one fligt. A significant differences is observed between and altitude. AHR2 Alt is a barometer altitude plus ome altitude (GPS altitude obtain in ome/start position), POS Alt is EKF Altitude output. During te fligt POS Alt is corrected. It is clearly visible on Fig. in te first pase of fligt, were EKF is correcting te POS Alt, based on IMU/GRO Z, BARO and GPS Alt. N.sats HDOP

3 Pos Positions in local reference system.5.5 Pos EKF Pos Ref. Pos Desired position 2 EKF Pos Ref. Pos Desired position Fig. 2. positions form different UAVs navigations system 0 Altitude 5 0 POS Alt AHR2 Alt Fig.. Altitude sensor readings differences and VDOP On board UAV, in loiter fly mode and axis location is GPS dependent, Z axis is BARO dependent. For direct poto georeferencing (Grejner-Brzezinska 2000) GPS location can be considered for and axis, and Z axis can be considered only form POS Alt, wic is EKF output for altitude. Tis is also important tat after a fast altitude canges EKF output POS Alt requires some time to calculate altitude value. Data acquisition Te field works were carried out on te Gdansk Tecnical University campus. A test fligt was planned and executed in good weater conditions and wit a wind speed of 1 to 2 m/s NE direction, temperature deg. C. Te GPS receiver of te UAV was operated on te ground before and after te fligt. Eac test fligt was carried out only after and fixed maximal number of satellites on tested position. Te UAV was cooled down to an air temperature, before test fligts. It remained minimum 0 minutes in tested environment. Te cooling down process is required to minimalize GRO and ACCEL bias. Te test field is considered as a not good GPS environment. fixed considerably lower number of satellites (Fig. ) during te test fligt. It as to be mentioned, tat in good environment of at te iger altitude used GPS modules can fix maximal 20 satellites.

4 NSats Numebr of sattelites HDOP 1.5 VDOP VDOP VDOP HDOP 1 VDOP a) b) Measurement system setup Fig.. Satellites reception during test fligt a) NSAT and HDOP, b) VDOP Te drone fligt data was logged on board fligt controller and on te ground station. Two syncronized cameras was triggered simultaneously wit RC cannel number seven (RC CH 7) (Fig. 5). Te ig PWM signal state of RC CH7 allowed to trigger cameras and syncronize wit FC system time (). Tis syncronization metod allowed to compare cannel state wit system time () and referenced potogrammetric position. Te reference drone position (Fig. 2) was obtained based on images from two syncronized cameras (Janowski et al. 2016; Janowski, Szulwic 201). Stereopairs were developed in te Image Master. Tis allowed te position measurement of te antenna. Te reference position referred to Poland 2000 coordinate system (zone 6) EPSG:2177. GPS logged position was recalculated to Poland 2000 coordinate system. Fig. 5. Measurement system setup Evaluation Standard deviation and accuracy measures for te presented UAV navigation system are calculated in accordance wit equations presented in (NovAtel 200) and results for eac system are presented in Table 2. Visualization D position accuracy measures for are presented in Fig. 6. Heigt (z axis) is a GPS altitude for all calculations.

5 Table 2. Te drone position navigation systems accuracy measures Accuracy Measures EKF pos δ δ δz SEP MRSE % Sperical Accuracy Standard % Sperical Accuracy Standard Sperical Error Probable = Mean Radial Sperical Error = % Sperical Accuracy Standard = % Sperical Accuracy Standard = Fig. 6. D position accuracy measures visualization Conclusions Te paper presents results of navigation accuracy of ceap navigation modules installed on board UAV exacopter platform. As te result sows, module mounted on small carbon mast presented a significantly better accuracy tan, mounted in close distant form EMI sources. Te offset between navigation systems was observed. Currently, fligt controller system cannot consider offsets between navigation systems. Tere is no possibility to calculate offsets during te fligt. It can be calculated during post processing. Due to te fact, tat te wind blew te drone, it could not reac desired position. N-NE wind blew drone to S-SW direction. Navigation system and INS recorded tat, owever te real drone position as been moved a bit more tan GPS and INS recorded. It can be solved in te future using navigation offsets and wind corrections. An intensive growt of commercial exploitation of low-cost UAVs, te accuracy of positioning using GPS tecniques sould be taken into account. Awareness of tis tecnology work is very important wen raids are carried out in urban areas wit a ig infrastructures. In te case of air raids on open areas, e.g. around te coastal strip for analysis cliffs (Szulwic et al. 20) or monitoring of sea vessels movement (Bobkowska 2016), ig navigation accuracy is not necessary. In suc a case, supervision raid, even witout a navigation module, troug observation drone, is possible. 5

6 A future work is planned for develop te presented analysis. A reference position can be obtained using metod in (Wierzbicki, Krasuski 2016) during an open area fligt. Area cosen for a future work sould be widely open for a GPS signal in order to carry out te off wit maximum number of satellites (20 satellites). Autors declare tat tere is no competing financial, professional or personal interests from oter parties, as far as presented researc is concerned. References Ardupilot P ArduPilot DEV. Bobkowska, K Analysis of te objects images on te sea using Dempster-Safer Teory, in 17t International Radar Symposium (IRS) May 2016, Krakow, Poland. IEEE: 1. Burdziakowski, P Lowcost exacopter autonomous platform for testing and developing potogrammetry tecnologies and intelligent navigation systems, in Environmental Engineering 10t International Conference April 2017, Vilnius. Grejner-Brzezinska, D Direct exterior orientation of airborne imagery wit GPS/INS system: performance analysis, Navigation, Journal of te Institute of Navigation 6(): ttps://doi.org/ /j tb021.x Janowski, A.; Nagrodzka-Godycka, K.; Szulwic, J.; Ziółkowski, P Remote sensing and potogrammetry tecniques in diagnostics of concretestructures, Computers and Concrete 18(): ttps://doi.org/ /cac Janowski, A.; Szulwic, J Syncronic digital stereopotograpy and potogrammetric analyses in monitoring te flow of liquids, in open cannels 9t International Conference Environmental Engineering 22 2 May 201, Vilnius, Lituania, selected papers. NovAtel GPS Position Accuracy Measures, Novatel Positioning Leadersip: 1 6 [online], [cited 5 January 2017]. Available from Internet: Szulwic, J.; Burdziakowski, P.; Janowski, A.; Przyborski, M.; Tysiac, P.; Wojtowicz, A.; Kolodkov, A.; Matysik, K.; Matysik, M. 20. Maritime Laser Scanning as te Source for Spatial Data, Polis Maritime Researc 22(): 9 1. ttps://doi.org/10./pomr TME-Electronic Transfer Multisort Elektronik Ltd. [online], [cited 5 January 2017]. Available from Internet ttp:// Wierzbicki, D.; Krasuski, K Determination te coordinates of te projection center in te digital aerial triangulation using data from Unmanned Aerial Veicle, Aparatura Badawcza i Dydaktyczna 21: