Sensor Fusion Algorithm by Complementary Filter for Attitude Estimation of Quadrotor with Low-cost IMU

TELKOMNIKA, Vol.16, No.2, April 2018, pp. 868~875
ISSN: 1693-6930, accredited A by DIKTI, Decree No: 58/DIKTI/Kep/2013
DOI: 10.12928/TELKOMNIKA.v16i2.9020  868
Received October 25, 2017; Revised February 22, 2018; Accepted March 10, 2018
Sensor Fusion Algorithm by Complementary Filter for
Attitude Estimation of Quadrotor with Low-cost IMU
A. Noordin
1
, M. A. M. Basri*
2
, Z. Mohamed
3
1
Department of Electrical Engineering Technology, Faculty of Engineering Technology,
Universiti Teknikal Malaysia Melaka, Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia
1,2,3
Department of Control and Mechatronics Engineering, Faculty of Electrical Engineering,
Universiti Teknologi Malaysia, 81310 Johor Bahru, Johor, Malaysia
*Corresponding author, e-mail: ariffanan@fke.utm.my
Abstract
This paper proposes a sensor fusion algorithm by complementary filter technique for attitude
estimation of quadrotor UAV using low-cost MEMS IMU. Angular rate from gyroscope tend to drift over a
time while accelerometer data is commonly effected with environmental noise. Therefore, high frequency
gyroscope signal and low frequency accelerometer signal is fused using complementary filter algorithm.
The complementary filter scaling factor K1=0.98 and K2=0.02 are used to merge both gyro and
accelerometer. The results show that the smooth roll, pitch and yaw attitude angle can be obtained from
the low cost IMU by using proposed sensor fusion algorithm.
Keywords: MEMS, complementary filter, sensor fusion, UAV, low-cost IMU
Copyright © 2018 Universitas Ahmad Dahlan. All rights reserved.
1. Introduction
Nowadays, advanced in technology innovates to combine mechanical and electrical
components into micro-scale object calls as Microelectromechanical systems (MEMS) [1], [2].
With MEMS technology component being inexpensive, small size, low power consumption,
therefore products are redesigned to include such as an inertial measurement unit (IMU) sensor
onto telecommunication, automotive industries and medical application [3-6]. In robotics, MEMS
technology are the main of attitude and heading reference system (AHRS) to determine rotation,
motion, location and direction (generally called attitude estimation) of mobile robots for
automated navigation, autonomous underwater vehicle (AUV) for global localization systems
and unmanned aerial vehicle (UAV) in aviation [1], [6-9].
However, MEMS sensor performance is commonly effected by biases and noises which
tend to drift over time and reduce the accuracy of measurement. Therefore, most researcher
applied sensor fusion algorithms techniques to overcome the measurement errors and obtaining
accurate reading [1], [6], [7], [10].
In present-day, nonlinear filter techniques for instance extended kalman filter (EKF) [1],
[11-13], unscented kalman filter (UKF) [14-17] and complementary filter (CF) [18-21] have been
developed, reformed and integrated to obtain the optimal fusion of sensors or in other term
called multi fusion integration (MFI) [22]-[23].
This paper focused on low cost IMU fusion using complementary filter for attitude
estimation of quadrotor UAV. The system utilizes APM2.6’s built in sensor (MPU6000) and
external HMC5833L. Data from both sensors are converted to angles before fusion using
complementary filter algorithm. The magnetic force vector from HMC5833L is used to determine
yaw angle. This complementary filter is used to overcome drift on gyroscope and reduce
environmental noise effect on accelerometer, and a smooth signal for attitude roll (ϕ) and pitch
(θ) estimator is expected.
This paper organized as follows. Section 1 briefly described quadrotor UAV
applications, mechanical architecture and dynamics movement, which in this research a X-
configuration quadrotor is used. Section 2 explained in general regarding sensors used in IMU
such as gyroscope, accelerometer, magnetometer and mention a sensor fusion algorithm
technique based on complementary filter used in this research. Section 3 explained the

TELKOMNIKA ISSN: 1693-6930 
Sensor Fusion Algorithm by Complementary Filter for Attitude Estimation of… (A. Noordin)
869
experiments arrangement and present the outcomes, and finally, Section 4 concludes the paper
by summarizing the findings.
1.1. Quadrotor UAV
Quadrotor is an aerial vehicle that has capabilities in vertical take-off and landing
(VTOL), omni-directional flying, and easy hovering performances in limited spaces. Compare to
terrain mobile robot, quadrotor can cover mission at most range due to capable of mobilizing in
environment with many obstacles. Furthermore, advances in MEMS technology [24-28] and
rapid prototyping technology makes quadrotor has a simplest electronics and mechanical
structures design. Therefore, quadrotor always being considered in research with wide
applications and various tasks such as civil commercial, calamity and meteorological
observation, agricultural and military purpose such as surveillances, reconnaissance and
communication [29-31] also educational purposes [32].
Quadrotor is a helicopter kind that can be controlled by varying the rotor speeds. It is an
under-actuated, dynamic vehicle with four input forces and six output coordinates. To eliminate
the anti-torque during manuaver, quadrotor configuration is setup as follows; 1) two diagonal
motors (1 and 2) are running in the same direction, 2) the others set of motors (3 and 4) in the
other direction [33-35].
1.2. Quadrotor Dynamics Movement
On varying the rotor speeds altogether with the same rate, the forces will enable VTOL
in altitude z of the quadrotor either hovering (by increasing speeds) as Figure 1a or landing (by
decreasing speeds) as Figure 1b. Manipulate roll angle, ϕ allows the quadrotor to move toward
y direction as shown in Figure 1c and 1d. Manipulate pitch angle, θ allows the quadrotor to
move toward x direction as shown in Figure 1e and 1f. To able the quadrotor to rotate in yaw
angle ψ direction, the diagonal motors need to be speed up/slow down depending on desired
direction as shown in Figure 1g and 1h [31], [34], [33].
(a) Vertical up (c) Moving right (e) Moving forward (g) Rotate right
(b) Vertical down (d) Moving left (f) Moving backward (h) Rotate left
Figure 1. Quadrotor dynamics manoeuvres
2. Research Method
With advanced in MEMS, combination of 3-axis gyroscopes and a 3-axis accelerometer
in one chip able to provide a full 6 degree-of-freedom (DOF) motion tracking system for many
applications [36]. With 3-axis digital compass additional of 3-DOF can assist on a heading for a

 ISSN: 1693-6930
TELKOMNIKA Vol. 16, No. 2, April 2018 : 868 – 875
870
system for correct orientation during locomotion. Figure 2 shows 6-DOF MPU6000 comprise of
gyro and accelerometer. 3-axis compass can be attached as auxiliary sensor for attitude
estimator system such used in APM2.6.
Figure 2. A 6-DOF–Gyro/Accelerometer
2.1. Accelerometer
An accelerometer sensor as shown in Figure 3 is a dynamic MEMS sensor used to
measure acceleration forces up to three orthogonal axes. Measurement unit of accelerometer
typically in gravitational motion (g) where 1g is equivalent to 9.8m/s2 which is acceleration
cause of earth gravity. However, for linear displacement measurement, recorded data need to
be integrated two times which tend to drift over a time and very sensitive to environmental
noise [5],[18].
Figure 3. A 3-Axis accelerometer sensor (source: Adafruit Industries, 2017)
In other way, from an accelerometer, a tilt angle can be sensed and estimated as
described on Freescale Semiconductor Application Note used by [37]. Figure 4a shows
technique using two axes from an accelerometer to measure tilt angle. As shown in Figure 4b, a
basic trigonometry can be applied to determine tilt angle (𝜃) as:
𝜃 = tan−1
(
𝐴𝑥
𝐴𝑦
) (1)
Where magnitude of acceleration is always 1g can be determined by:
𝐴 = √ 𝐴 𝑥
2 + 𝐴 𝑦
2 = 1𝑔 (2)
a) Tilt Angle Sensing (b) Rotate about y-axis
Figure 5. Three axes tilt sensing for (a) roll angle and (b)
pitch angle

871
2.2. Gyroscope
A gyroscope is a device typically used for navigation by measurement of angular
velocity and estimated changing in orientation [38]. However, for a long term, gyroscope reading
tends to drift from static condition. Gyroscopes can measure rotational velocity up to three
directions. In IMU, a gyroscope or gyros sensor as shown in Figure 6 is used to detect rotation
speed to measure angular velocity or in other term called angular rate with measurement unit
degree per second (°/s) or revolution per second (RPS).
Figure 6. A 3-axis gyro sensor (Source: Adafruit Industries, 2017)
By integrating the angular rate (reading from sensor) of each axis, the roll angle (ϕ), the pitch
angle (θ), and the yaw angle (ψ) can be obtained using Equation 5:
𝜙 𝑔𝑦 = ∫ 𝜔 𝑥 𝑑𝑡
𝜃𝑔𝑦 = ∫ 𝜔 𝑦 𝑑𝑡
𝜓 𝑔𝑦 = ∫ 𝜔𝑧 𝑑𝑡 (5)
Standard control range of angle in aviation system is between -180 to 180 degrees. Therefore,
result from Equation 5 need to be adjusted within this range.
2.3. Magnetometer
A magnetometer as shown in Figure 7 can sense where the strongest magnetic force is
coming from, generally used to detect magnetic north. Therefore, as for digital compass,
magnetometer sensor is used to provide heading information where the heading system
calculation depending to magnetic data (mx, my, mz) therefore the compass orientation can be
mathematically rotated to horizontal plane as shown in Figure 8. Based on this figure, mx, my,
and mz from magnetic sensor reading can be transformed to the horizontal plane (Xh, Yh) by
applying the rotation as Equation 6 and Equation 7 with aided from tilt sensing roll (𝜙) and pitch
(𝜃) from an accelerometer sensor. As such it is useful to determine absolute orientation in the
north-east-west-south (NEWS) plane of any systems during navigation [37].
Figure 7. A 3-Axis magnetometer sensor
(source: adafruit industries, 2017)
Figure 8. Tilt sensing angle in local horizontal
plane defined by gravity (caruso, 2000)
𝑋ℎ = mx cos(𝜙 𝑎𝑐𝑐) + my sin(𝜙 𝑎𝑐𝑐) − mz cos(𝜃 𝑎𝑐𝑐) sin(𝜙 𝑎𝑐𝑐) (6)
𝑌ℎ = my cos(𝜃 𝑎𝑐𝑐) + mzsin(𝜃 𝑎𝑐𝑐) (7)

 ISSN: 1693-6930
872
From Equation 6 and Equation 7 then, the heading information can be calculated as:
𝐻𝑒𝑎𝑑𝑖𝑛𝑔 = 𝜓ℎ = tan−1
(
𝑌ℎ
𝑋ℎ
) (8)
2.4. Sensor Fusion Algorithm
There are two ways to generate attitude estimation for quadrotor; from gyroscopes by
integrating it values or from accelerometer vector and magnetometer vector. Attitude estimation
using gyroscope provides a good dynamics performance. While fusion using accelerometer
vector and magnetometer vector give better steady-state performances [18].
In this attitude estimation, a fusion by high frequency gyroscope signal and low
frequency accelerometer signal and magnetometer signal is used. This technique called
complementary filter (CF) used by most researcher in [18], [20], [21]. This fusion provides
reliable roll (ϕ) and pitch (θ) reading from gyroscope and accelerometer. While fusion tilt sensing
roll (ϕ) and pitch (θ) from accelerometer with magnetic vector, (mx, my, mz) provide yaw (ψ)
which is heading direction of the system [20]. Figure 9 shows a block diagram to illustrate
proposed CF technique used in this attitude estimation.
Figure 9. Complementary filter for attitude estimation
In complementary filter system, the angles will be updated with new gyroscope
measurement by integral (summing) over a time and tilt sensing from accelerometer
measurement with scaling factor K. The mathematic expressions to represent complementary
filter used in this study
𝜙 𝐶𝐹 = 𝐾1(𝜙 𝐶𝐹 + 𝜙 𝑔𝑦) + 𝐾2 𝜙 𝑎𝑐𝑐 (9)
𝜃 𝐶𝐹 = 𝐾1(𝜃 𝐶𝐹 + 𝜃𝑔𝑦) + 𝐾2 𝜃 𝑎𝑐𝑐 (10)
𝐾1 + 𝐾2 = 1 (11)
3. Results and Analysis
Gyroscope, accelerometer, and magnetometer data from APM2.6 System is recorded
and captured by serial oscilloscope. The system is program to record sensors data with
sampling period of 50Hz for gyroscope and accelerometer while 10Hz for magnetometer. Figure
10 shows the measurement of accelerometer and gyroscope after scaling but before filtering.
Plotted angle of gyroscope shows a tend to drift over a time, while accelerometer angle very
sensitive to noise. In static condition both angles supposed to be at zero angle. Hence, a
complementary filter is utilized to overcome the interference.

873
Figure 10. Gyroscope and accelerometer data converted to angle
Figure 11a and 11b show comparisons for roll angle (ϕ) and pitch angle (θ) respectively
between gyroscope, accelerometer and after complementary filter is implemented. For Figure
11a, the spikes at time 4s and 8.4s are roll angle recorded when the systems are manually
rotate for positive roll angle and negative roll angle as expected the quadrotor will move to the
right and left (refer to Figure 1c and 1d respectively). While in Figure 11b, spike at 6.8s and 9.2s
are pitch angle during quadrotor move forward and backward (refer to Figure 1e and 1f
respectively). All rotation angles follow right hand rule orientation.
(a) roll (𝜙) angle (b) pitch (𝜃) angle
Figure 11. Real-time roll (ϕ) angle measurement and pitch (θ) angle measurement
From both graphs, gyro roll (𝜙) and pitch (𝜃) signal drift over a time and not return to
zero in static condition due to integral (summation) calculated from gyroscope angle reading.
Accelerator roll (𝜙) and pitch (𝜃) signal from both graphs consist of noise as can be seen in
Figure 10 since low cost MEMS sensor sensitive to environment. The complementary filter
signal for both graph which is the estimated roll and pitch angle are the combine of gyro and
accelerometer with scaling factor 𝐾1 = 0.98 and 𝐾2 = 0.02.
4. Conclusion
In this work, a complementary filter was implemented with low cost IMU to obtain
attitude estimation for quadrotor system. 3-axis gyroscope and 3-axis accelerometer in
MPU6000 of APM2.6 is recorded for angular rate by integrating gyros data and tilt sensing from
accelerometer. Angular rate from gyros tend to drift over a time due to summing function
(integral) while accelerometer is commonly effected to environmental noise. Hence a
complementary filter algorithm is applied to update attitude roll (𝜙) and pitch (𝜃) by merging
angular rate and tilt sensing data with appropriate scaling factor. Experiment result shows

 ISSN: 1693-6930
874
complementary filter technique can overcome drift problem by gyroscope which is easy to
implement for quadrotor attitude estimation of low cost IMU.
Acknowledgements
The authors would like to thank Universiti Teknologi Malaysia (UTM) under the
Fundamental Research Grant Scheme (R.J130000.7823.4F761), Research University Grant
(Q.J130000.2523.15H39), Universiti Teknikal Malaysia Melaka (UTeM), and Ministry of Higher
Education for supporting this research.
References
[1] Y Hu, Y Yan, J Liang, L Wang. A Miniature, Low-cost MEMS AHRS with Application to Posture
Control of Robotic Fish. Conf. Rec.-IEEE Instrum. Meas. Technol. Conf. 2013: 1392–1395
[2] B Allotta, R Costanzi, F Fanelli. An Attitude Estimation Algorithm for Underwater Mobile Robots.
IEEE/ASME Trans. Mechatronics. 2016; 21(4): 1900–1911.
[3] DR Santoso, S Maryanto, A Nadhir. Application of Single MEMS-Accelerometer to Measure 3-Axis
Vibrations and 2-Axis Tilt-Angle Simultaneously. TELKOMNIKA (Telecommunication, Computing,
Electronics and Control) . 2015; 13(2): 442–450.
[4] K Badshah, Q Yongyuan. Integration and Simulations of INS/GNSS System Using the Approach of
Carrier Phase Measurements. Int. J. Robot. Autom. 2015; 4 (4): 243–253.
[5] U Guner, H Canbolat, A Unluturk. Design and Implementation of adaptive vibration filter for MEMS
based low cost IMU. 2015 9th Int. Conf. Electr. Electron. Eng. 2015; 130–134.
[6] W Lingling, F Li, H Xioguang, Z Guofeng. Attitude Estimation for UAV with Low-Cost IMU/ADS Based
on Adaptive-Gain Complementary Filter. 2016; 54: 673–680.
[7] Z Wu, Z Sun, W Zhang, Q Chen. Attitude and gyro bias estimation by the rotation of an inertial
measurement unit. Meas. Sci. Technol. 2015; 26(12): 125102.
[8] DQ Duong, J Sun, TP Nguyen, L Luo. Attitude estimation by using MEMS IMU with Fuzzy Tuned
Complementary Filter. Proc. 2016 IEEE Int. Conf. Electron. Inf. Commun. Technol. ICEICT 2016, no.
Iceict. 2017: 372–378.
[9] BH Prasetio, AA Supianto, GE Setiawan. Earth Image Classification Design Using Unmanned Aerial
Vehicle. TELKOMNIKA (Telecommunication, Computing, Electronics and Control). 2015; 13(3): 1021–
1028.
[10] Y Wang, N Li, X Chen, M Liu. Design and implementation of an AHRS based on MEMS sensors and
complementary filtering. Adv. Mech. Eng. 2014.
[11] S Sabatelli, M Galgani, L Fanucci, A Rocchi. A double-stage kalman filter for orientation tracking with
an integrated processor in 9-D IMU. IEEE Trans. Instrum. Meas. 2013; 62(3): 590–598.
[12] D Veillard, F Mailly, P Fraisse. EKF-based state estimation for train localization. Proc. IEEE Sensors.
2017; 2(2): 5–7.
[13] P Yutthaa, J Tatsawan, P Chotikunnan, T Nuntachi, L Wanichsara, T Atimon. The Effect of Average
Filter for Complementary Filter and Kalman Filter Based on Measurement Angle. Biomed. Eng. Int.
Conf. 2016: 1–4.
[14] SA Maged, AA Abouelsoud, AMRF El Bab, T Namerikawa. Stewart Platform Manipulator: State
Estimation Using Inertia Sensors and Unscented Kalman Filter. Proc-2016 3rd Int. Conf. Inf. Sci.
Control Eng. ICISCE 2016: 1136–1140.
[15] WU Zhonghong, SHI Zhangsong, LIU Jian. Multi-sensor Fusion Based Unscented Attitude Estimation
Method for MAVs. 2016: 5138–5142.
[16] D Yuan, X Ma, Y Liu, C Zhang. Dynamic initial alignment of the MEMS-based low-cost SINS for AUV
based on unscented Kalman filter. Ocean. Shanghai, 2016.
[17] Z Lin, X Linlin, L Fanming, C Yi. Application of UKF for MEMS IMUs and fluxgate sensors based
attitude and heading reference system of carriers. ICIEA 2007 2007 Second IEEE Conf. Ind. Electron.
Appl. 2007; 0: 2278–2283.
[18] NH Ariffin, N Arsad, B Bais. Low cost MEMS gyroscope and accelerometer implementation without
Kalman Filter for angle estimation. Int. Conf. Adv. Electr. Electron. Syst. Eng. 2016: 77–82.
[19] IMUW Kalman, TC Filter. Reading a IMU without Kalman : The Complementary Filter. 2013: 1–9.
[20] X Lu, J Li, W Liu. Study of Complementary Filter Algorithm for Four-rotor Helicopters Attitude Control
System. 2014; 174(7): 298–306.
[21] O De Silva, GKI Mann, RG Gosine. Automated tuning of the nonlinear complementary filter for an
Attitude Heading Reference observer. IEEE Work. Robot Vision, WORV. 2013: 171–176.
[22] M Faiz, B Ramli, SS Shamsudin, A Legowo. Obstacle Detection Technique Using Multi Sensor
Integration for Small Unmanned Aerial Vehicle. Indonesian Journal of Electrical Engineering and
Computer Science (IJEECS). 2017; 8(2): 441–449.

875
[23] RC Luo, CC Chang, CC Lai. Multisensor fusion and integration: Theories, applications, and its
perspectives. IEEE Sens. J. 2011; 11(12): 3122–3138.
[24] A Cahyadi, S Herdjunanto, R Hidayat. Inertial Navigation for Quadrotor Using Kalman Filter with Drift
Compensation. International Journal of Electrical and Computer Engineering (IJECE). 2017; 7(5):
2596–2604.
[25] J Chang, J Cieslak, A Zolghadri, J Dávila, J Zhou. Quadrotor attitude estimation with gyroscope bias
reconstruction capabilities. IFAC-PapersOnLine. 2016; 49(5): 260–265.
[26] V Kriz, P Gabrlik. UranusLink-Communication protocol for UAV with small overhead and encryption
ability. in IFAC Proceedings Volumes (IFAC-PapersOnline). 2015; 48(4): 474–479.
[27] PJ Bristeau, F Callou, D Vissière, N Petit. The Navigation and Control technology inside the AR.Drone
micro UAV. in IFAC Proceedings Volumes (IFAC-PapersOnline). 2011; 18(PART 1): 1477–1484.
[28] H Jafari, M Zareh, J Roshanian, A Nikkhah. An Optimal Guidance Law Applied to Quadrotor Using
LQR Method. Trans. Jpn. Soc. Aeronaut. Space Sci. 2010; 53(179): 32–39.
[29] J Ajmera, V Sankaranarayanan. Point-to-point control of a quadrotor: Theory and experiment. IFAC-
PapersOnLine. 2016; 49(1): 401–406.
[30] J Bazin, T Fields, AJ Smith. Feasibility of In-Flight Quadrotor Individual Motor Thrust Measurements.
AIAA Atmos. Flight Mech. Conf., no. January 2016: 1–12.
[31] DBBVL, P Singh. A survey on design and development of an unmanned aerial vehicle (quadcopter).
Int. J. Intell. Unmanned Syst. 2016; 4(2): 70–106.
[32] J Števek, M Fikar. Teaching Aids for Laboratory Experiments with AR.Drone2 Quadrotor**The authors
gratefully acknowledge the financial support of the Slovak Research and Development Agency under
the project APVV 0551-11 and the Scientific Grant Agency of the Slovak Republic. IFAC-
PapersOnLine. 2016; 49(6): 236–241.
[33] SK Phang, C Cai, BM Chen, TH Lee. Design and mathematical modeling of a 4-standard-propeller
(4SP) quadrotor. in Proceedings of the World Congress on Intelligent Control and Automation
(WCICA). 2012: 3270–3275.
[34] D Domingos, G Camargo, F Gomide. Autonomous Fuzzy Control and Navigation of Quadcopters.
IFAC-PapersOnLine. 2016; 49(5): 73–78.
[35] S Bouabdallah, P Murrieri, R Siegwart, B Ghder, PR Bud. Design and Control of an Indoor Micro
Quadrotor. Proc. 2004 IEEE Int. Conf. Robot. 8 Autom. New Orleans, LA April 2004: 4393–4398.
[36] S Amin, M Bin, I Reaz, SS Nasir. Integrated Vehicle Accident Detection and Location System.
TELKOMNIKA (Telecommunication, Computing, Electronics and Control). 2014; 12(1): 73–78.
[37] D Derawi, J Kim. Real-time Nonlinear Complementary Observer for Low-cost Inertial Attitude System.
Int. Glob. Navig. Satell. Syst. Soc. IGNSS Symp. 2013.
[38] H Fourati, N Manamanni, L Afilal, Y Handrich. Complementary observer for body segments motion
capturing by inertial and magnetic sensors. IEEE/ASME Trans. Mechatronics. 2014; 19(1): 149–157.

Sensor Fusion Algorithm by Complementary Filter for Attitude Estimation of Quadrotor with Low-cost IMU

More Related Content

What's hot (20)

Similar to Sensor Fusion Algorithm by Complementary Filter for Attitude Estimation of Quadrotor with Low-cost IMU (20)

More from TELKOMNIKA JOURNAL (20)

Recently uploaded (20)

Sensor Fusion Algorithm by Complementary Filter for Attitude Estimation of Quadrotor with Low-cost IMU